COMPARATIVE RISKS OF HAZARDOUS MATERIALS AND
NON-HAZARDOUS MATERIALS TRUCK SHIPMENT ACCIDENTS/INCIDENTS

Final Report

Prepared for
Federal Motor Carrier Safety Administration

March 2001

505 King Avenue, Columbus, OH 43201

Table of Contents	Page
Executive Summary	ES-1
1.1 Purpose and Organization	1-1
1.2 Hazardous Materials Transportation	1-2
1.3 Hazardous Material Flow	1-3
2.0 Study Methodology	2-1
2.1 Accident and Incident Data Sources	2-1
2.2 Methodology for Estimating Accidents/Incidents	2-5
2.3 Impact Methodology	2-10
2.4 Impact Summary Discussion	2-25
3.0 High Consequence Low Probability Accidents	3-1
3.1 Introduction	3-1
3.2 Event Tree Application	3-1
4.0 Accident and Incident Numbers and Impacts	4-1
4.2 Shipment Impact Summary	4-2
5.0 Impact Summary by HM Category	5-1
5.1 Introduction	5-1
5.2 HM Category 1.1, 1.2, 1.3	5-1
5.3 HM Category 1.4, 1.5, 1.6	5-1
5.4 HM Category 2.1	5-1
5.5 HM Category 2.2	5-9
5.6 HM Category 2.3	5-9
5.7 HM Category 3	5-9
5.8 HM Category 4.1, 4.2, 4.3	5-10
5.9 HM Category 5.1, 5.2	5-10
5.10 HM Category 6.1, 6.2	5-10
5.11 HM Category 7	5-10
5.12 HM Category 8	5-10
5.13 HM Category 9	5-11
5.14 All HM Categories	5-11
6.0 Non-Hazardous Materials Accident Totals and Impacts	6-1
6.1 Introduction	6-1
6.2 Accident and Incident Totals	6-1
6.3 Non-HM Accident Impacts	6-1
6.4 Non-HM Accident Risk and Cost per Mile	6-2
7.0 Comparative Impacts and Risk of HM and Non-HM Shipments	7-1
7.1 Introduction	7-1
7.2 Comparative Costs	7-1
7.3 Comparative Risk and Cost per Mile	7-1
7.4 Discussion	7-4
8.0 SafeStat Applications	8-1
8.1 Introduction to SafeStat	8-1
8.2 Current Role of HM in the SafeStat Algorithm	8-1
8.3 SafeStat HM Analysis	8-2
8.4 Changes in SafeStat Applications	8-4
9.0 HM Database Assessments and Recommendations	9-1
9.1 Purpose and Organization	9-1
9.2 Opportunities for Database Improvement	9-2
9.3 Recommendations and Conclusions	9-5
10.0 Conclusion	10-1
10.1 Project Significance	10-1
10.2 Project Results	10-2
10.3 Recommendations for Future Projects	10-3

List of Appendices	Page
Appendix A. State Hazardous Material Flows	A-1
Appendix B Database Search Criteria	B-1
Appendix C. Accident Information for Class 3, Class 2.1, and Class 8	C-1
Appendix D Natural Resource Damages Settlements	D-1
Appendix E Class 2.1 Releases	E-1
Appendix F Impact Case Examples	F-1
Appendix G References	G-1

List of Tables	Page
Table 1. Class 3 Truck Shipments — Estimated Unique Accidents for 1996	2-9
Table 2. Class 2.1 Truck Shipments — Estimated Unique Accidents for 1995–1997	2-9
Table 3. Class 8 Truck Shipments — Estimated Unique Accidents for 1995–1997	2-10
Table 4. Distribution of VMT by Functional Class and Vehicle Type	2-18
Table 5. Summary of Steps	2-22
Table 6. Initiating Event Frequency and Event Tree Branch Probabilities for Various Classes/Divisions of HM Being Transported by Truck	3-9
Table 7. Initiating Event Frequency and Event Tree Branch Probabilitiesfor Division 2.3 (Poison Gases) by Truck	3-10
Table 8. Initiating Event Frequency and Event Tree Branch Probabilitiesfor Class 7 (Radioactive) by Truck	3-10
Table 9. HM Accident and Incident Likelihood	4-1
Table 10. Enroute Release Accident Types	4-2
Table 11. Estimated Annual Accident and Incident Impacts (Costs)	4-3
Table 12. Estimated Annual Release Accident Impact Costs	4-3
Table 13. Average Accident/Incident Costs for the Portrait Year	4-5
Table 14. Average Accident Costs for the Portrait Year	4-5
Table 15. Average Impacts per Accident as if the High Consequence/Infrequent Accident had Occurred in the Portrait Year Compared to Average Costs for that Year	4-6
Table 16. Enroute Accidents: Total HM (Release/Non-Release) Impact Components	4-7
Table 17. Enroute Accident: Total Releases, Impact Components	4-9
Table 18. Enroute Accident: Release Only, Impacts Components	4-9
Table 19. Enroute Accident: Fire, Impact Components	4-10
Table 20. Enroute Accident: Explosion Impact Components	4-10
Table 21. Enroute Accidents Non-Release Accidents Impact Components	4-11
Table 22. Leak Enroute Impact Components	4-11
Table 23. Loading/Unloading Incidents	4-12
Table 24. HM Accident Rate Per Mile	4-13
Table 25. HM Accident/Incident Risk Per Mile (Includes Leak Enroute Incidents)	4-13
Table 26. HM Accident Cost Per Mile	4-14
Table 27. HM Accident/Incident Cost Per Mile (Includes Leak Enroute)	4-14
Table 28. Category 1.1, 1.2, 1.3	5-2
Table 29. Category 1.4, 1.5, 1.6	5-2
Table 30. Category 2.1	5-3
Table 31. Category 2.2	5-3
Table 32. Category 2.3	5-4
Table 33. Category 3.0	5-4
Table 34. Category 4.1, 4.2, 4.3	5-5
Table 35. Category 5.1, 5.2 .	5-5
Table 36. Category 6.1, 6.2	5-6
Table 37. Category 7	5-6
Table 38. Category 8	5-7
Table 39. Category 9	5-7
Table 40. All HM Categories	5-8
Table 41. Estimated Non-HM Accidents in 1996	6-1
Table 42. Estimated Annual Non-HM Accident Impacts	6-2
Table 43. HM and Non-HM Accident Rate per Mile	7-2
Table 44. HM Accident and Non-HM Accident Cost per Mile	7-4
Table 45. Effect of Removing All HM Weightings from the SafeStat Scoring Algorithm	8-7
Table 46. Effect of Adding Non-Spill Accidents to the Accident SEA Algorithm	8-8
Table 47. Effect of Doubling Weight of Spill and Non-Spill Accidents to the Accident SEA Algorithm	8-8
Table 48. Effect of Scoring all Bulk HM with Accident SEA >70	8-10
Table 49. Effect of Including Bulk Carriers with ACSEA Scores >70 and Adding Non-Spill to the Spill Accidents and Doubling the Weighting	8-10

COMPARATIVE RISKS OF HAZARDOUS MATERIALS AND
NON-HAZARDOUS MATERIALS TRUCK SHIPMENT
ACCIDENTS/INCIDENTS

Executive Summary

This project was designed to assist the U.S. Department of Transportation (USDOT) in
achieving their strategic goal of reducing the rate and severity of transportation fatalities
and injuries in hazardous materials (HM) transportation and of reducing the dollar loss from
high-consequence, transportation accidents. The purpose of this project is to assess the additional
risks posed by HM transportation when compared to non-hazardous shipments. The results will
also be used to assist the Federal Motor Carrier Safety Administration (FMCSA) in identifying high
risk motor carriers.

The project has been divided in three phases.

· The initial portion of the first phase characterized for one year the shipment impacts of Class 3
shipments and assessed the feasibility of conducting a comprehensive risk assessment of HM
and non-HM shipments. The feasibility study results were published in the Plan for Assessing
the Feasibility for Conducting a Comparative Risk Assessment on Hazardous Materials and
Non-hazardous Materials Movements, June 1999.

· The second part of the first phase characterized the one-year shipment impacts of Class 2.1 and
Class 8 and provided a preliminary estimate of the impacts of non-HM shipments. Two white
papers, Potential for Integrating Hazmat Transportation Risk Assessment into Safestat and
Incorporating Severe Class 3 and Class 2.1 Accidents into the Truck Transportation Risk
Assessment were produced by the project in 1999.

· The project’s second phase was the actual comparative risk assessment between HM and non-
HM truck shipments. To obtain the overall HM risk, the study calculated the risk associated
with each class/division of hazardous material. With the completion of the second phase of the
project, the risk associated with the shipment of any class/division of HM can be compared to
the risk associated with other classes/divisions as well as to the shipment of non-HM materials.

· The third phase of the project focused on a possible application of the HM risk results.
Specifically, the study assessed how HM risk information is currently being used in the SafeStat
program to identify potentially unsafe HM carriers. With these results in hand, the assessment
then focused on how the HM risk information obtained during the first two phases of the study
could be applied to the SafeStat algorithm to better identify “high risk” HM carriers.

· For comparative purposes, the risk assigned was applied to 12 HM categories, consisting of
classes and divisions or groups of divisions as follows:

· Class 1: Divisions 1.1, 1.2, 1.3 — all have the potential for mass detonation

· Class 1: Divisions 1.4, 1.5, 1.6 — characteristics make mass detonation unlikely

· Class 2: Division 2.1 — Flammable gases

· Class 2: Division 2.2 — Non-flammable gases

· Class 2: Division 2.3 — Poisonous gases

· Class 3: Flammable liquids (and combustible liquids)

· Class 4: Division 4.1, 4.2, 4.3— Flammable solids; spontaneously combustible materials
and dangerous when wet materials

· Class 5: Division 5.1, 5.2 — Oxidizers and organic peroxides

· Class 6: Division 6.1, 6.2— Toxic (poison) materials and infectious substances

· Class 7: Radioactive materials

· Class 8: Corrosive materials

· Class 9: Miscellaneous dangerous goods.

Adding non-HM transport brought the total number of categories of materials being assessed to 13.

This report analyzes events involving the transportation of hazardous material that may or may not
result in the release of hazardous material to the environment. These events are defined as accidents
and incidents. An accident is defined here as an event that occurs when the vehicle transporting the
goods is involved in a collision. Any accident involving the shipment of HM would be considered
as a HM accident regardless of whether any of the material was spilled or was exposed to the
atmosphere. Similarly, a non-HM shipment accident would be considered as a non-HM shipment
accident even if fuel from the tractor spilled during an accident. An event that occurs when the
vehicle transporting the goods spills some of the HM cargo but is not involved in a collision is
termed an enroute incident. An event resulting in the spill or release of HM material during loading
or unloading is defined as a loading/unloading incident.

An initial step in developing a risk assessment is to estimate reliably the number of accidents and
incidents across a defined period of time. For the first phase, estimates were developed for Class 3,
Division 2.1, and Class 8 truck shipment incidents and accidents for a representative year. The
Hazardous Materials Information System (HMIS) database served as the baseline database. The
HMIS represents the only national database of hazardous materials, highway transportation
incidents with details of the material, packaging and consequences involved. To be more complete,
the data found in the HMIS were supplemented with data from other federal and state databases.
The most important of these was the Motor Carrier Management Information System (MCMIS)
accident database that provides accident information for both spill and no-spill accidents. The study
determined underreporting rates for Class 3, Division 2.1, and Class 8 accidents and incidents by
examining the same accident in several databases. These underreporting factors were then applied
to the other HM categories to develop accident and incident likelihoods for an annual portrait.

A key portion of this assessment was the consideration of the impacts of high consequence/low
frequency accidents. First, these severe accidents were identified through an examination of the
historical record during the past fifty years. Next, the study obtained the likelihood of occurrence
by estimating the fraction of the accidents represented by the accident sequence that would, based
on the historical record, likely to be severe. Thus, a total likelihood of accidents for the portrait year
was developed for all of the HM categories.

The average annual enroute HM accident frequency was estimated to be 2,484 accidents. The
release accidents are estimated at 768. Average annual enroute leak incidents totaled 1,455 and
loading/unloading incidents totaled 10,746.

Class 3 shipments account for about 64 percent of the enroute accidents with releases and about 52
percent of the non-release accidents. Class 3 shipments along with categories 2.1, 2.2, 5.1, 5.2, 8,
and 9, represent about 94 percent of all enroute accidents with releases and about 93 percent of all
enroute non-release accidents.

Classes 3 and 8 alone are involved in about 77 percent of all of the enroute leaks in the year. For
loading and unloading incidents, these two classes were involved in about 84 percent of all
incidents.

To derive an estimate of the economic impact of incidents/accidents for the annual portrait, the
following impact categories were considered:

· Injuries and Deaths
· Cleanup Costs
· Property Damage
· Evacuation
· Product Loss
· Traffic Incident Delay
· Environmental Damage.

The study reviewed several sources of information to establish reasonable estimates of the
economic impacts of each consequence. A literature review was conducted, as was an evaluation
of the utility of the federal and state databases. Impact estimates not readily available from the
above sources, such as incident delay, were modeled. Finally, all impacts were converted to dollars
to permit comparison and to compile total impact cost.
The HMIS proved to be an important source of impact costs for product loss, cleanup costs, and
property damage. Injuries and deaths were valued to be the amount the USDOT would be willing
to spend to avoid an injury or death. This averaged out to be $200,000 to avoid an injury and
$2,800,000 to avoid a fatality.

Traffic incident delay was established as the total number of people delayed at an incident or
accident multiplied by $15 per hour. The size of an average spill and the value placed on
environmental contamination as determined by an average of 30 legal settlements constituted an
estimate of environmental damage.

Total HM annual impacts for the portrait year are estimated at about $1.2 billion. Enroute accidents
with total impacts of just over $1 billion account for about 89 percent of the total impacts.
Accidents with a release of HM with impacts of $416 million account for a total of about 40 percent
of the enroute accident impacts. Within the release accident category, accidents with a fire and
accidents with an explosion have total impacts of $139 million or about 34 percent of the total cost
of enroute release accidents. The consequences of these accidents are important because they make
up only about 12 percent of the total number of enroute release accidents. Non-release accidents
make up about 60 percent of the total enroute accident impacts in the annual portrait.

Leaks enroute at $72 million account for an additional six percent and loading/unloading incidents
at $53.5 million accounts for about five percent of the impacts.

Class 3 represents 56 percent of all of the impacts, while categories 8, 2.1, 2.2, and 9 represent 13
percent, 9 percent, 6 percent, and 7 percent respectively. These five categories alone account for
approximately 91 percent of the estimated annual impacts for HM shipments. No other category
accounts for more than three percent of the total impacts.

Injuries and fatalities dominated the impact costs. For both release and non-release accidents
combined, injuries represent about 40 percent of the impact costs. Fatalities represent about 40
percent of all impact costs for enroute accidents. Thus, injuries and fatalities together account for
about 80 percent of the impact cost. Incident delay for both release and non-release enroute
accidents add up to about nine percent of the total cost. Carrier, property damage, and product loss
together represent about eight percent of the total. Clean up, environmental damage, and
evacuations account for the remaining three percent of impacts.

Non-HM shipments experienced an estimated 126,880 accidents in the portrait year. After
compensating for underreporting, there were an estimated 5,009 fatalities and 109,779 injuries.
These injuries and fatalities result in impact costs of about $43 billion. All but $7 billion of that
cost results from injuries and fatalities.

All release and non-release enroute accidents for all of the HM categories for the annual portrait
year have an average value of about $414,000 per accident, while non-HM accidents averaged
about $340,000 per accident. This difference is magnified when non-HM accident impacts are
compared with HM release impacts. In the annual portrait year, the average cost per HM accident
release is about $536,000. The average impact cost of a release accident with a fire or one with an
explosion compared to the average cost of a non-HM accident shows an even greater contrast.

The non-HM accident rate of 0.73 per million vehicle miles is more than double the average HM
accident rate of 0.32 per million vehicle miles. This comparison is based on estimated mileage
figures from the 1997 Commodity Flow Survey (CFS). As stated above, the annual economic
impact of non-HM truck accidents is over $43 billion, considerably higher than for HM truck
incidents. Although due primarily to a much larger volume of transport activity, the estimated non-
HM truck accident rate is also reflected in the impact cost per vehicle-mile.

Hazardous material shipments make up between four and eight percent of all shipments. Given this
small percentage, the cost of non-HM accidents clearly dominates the cost of HM accidents.
Although the average cost of an accident is higher for HM, these higher costs are not nearly enough
to overcome the large disparity in shipment volume between HM and non-HM shipments by truck.

Taking these observations into consideration, one should view the results of this risk assessment in
the context of establishing a general estimate or bound on the financial impact of this problem
rather than a precise valuation. This project represents a systematic attempt to benchmark the
financial implications of the problem based on the best available data. We anticipate that
meaningful research and policy inferences can be derived for risk management purposes.

The SafeStat algorithm was evaluated to determine the appropriate inclusion of the risk of
hazardous materials shipments in the FMCSA carrier selection process. Potential changes in how
HM is used in the algorithm were the focus of this effort. At the present time, about 1.6 percent of
the bulk HM carriers are identified as potentially unsafe carriers and are therefore subjected to a
compliance review. However, the risk assessment results show that the cost associated with the
transport of bulk HM by truck represents over two percent of the total truck accident risk. Thus, the
current SafeStat algorithm under represents bulk HM carriers. Several alternative scenarios for
increasing this percentage were subsequently defined and evaluated. Based on these results, the
recommendations formulated state that all bulk HM carriers with a D score should undergo a
compliance review. In addition, the scoring algorithm should be changed for bulk HM carriers to
include all ACSEA scores greater than 70. Finally, the accident weighting for HM accidents should
be expanded to include both spill and non-spill accidents. Currently, SafeStat uses only HM spill
accidents in the accident weighting.

The HM risk assessment results presented in this study made extensive use of DOT, Census Bureau
and State supported databases. While these results would not be possible without the availability of
these databases, limitations of the study can in part be linked to their deficiencies. The study
concludes with recommendations, such as investigating ways to cross-reference the TIFA, MCMIS,
and HMIS databases and determining the causes of HM accidents. These would enable FMSCA to
improve its safety performance monitoring capabilities. The benefit of such improvements would
be a reduction in the expense associated with maintaining the databases and in the availability of
additional information, such as causal factors, that could be used to develop programs to improve
the safety of both HM and non-HM truck transport.

1.0 Introduction
1.1 Purpose and Organization

The United States Department of Transportation’s (U.S. DOT) 1997 Draft Strategic Plan
recognizes safety as its most important strategic goal and commits to promoting the public
health and safety by working towards the elimination of transportation related deaths,
injuries, and property damage. This project was designed to assist DOT in achieving this strategic
goal by reducing the rate and severity of transportation fatalities and injuries in hazardous materials
transportation and the dollar loss from high-consequence transportation accidents. Additionally, the
FMCSA 2000-2001 Hazardous Materials Program Plan stresses the identification of high risk
carriers for compliance reviews as a primary strategy for the reduction of hazardous materials
incidents. The information developed in this project will be directed toward that strategy.

The long-term purpose of this project is to assess the additional risks posed by hazardous materials
(HM) highway truck shipments when compared to non-hazardous materials (non-HM) highway
truck shipments. Specifically, the project focuses on benchmarking the risk associated with HM
highway transportation as compared to the transportation of non-HM. A second purpose of the
project is to develop a transportation risk assessment model that will enable the Federal Motor
Carrier Safety Administration (FMCSA) to identify programs that can result in the greatest
improvement in safety. Additionally, the FMSCA must be able to break down the HM risk
assessment into hazard classes so that experts can compare the costs associated with accidents/
incidents for each class. The distinction among hazard classes is based on the regulatory hazard
classification system that includes nine classes with divisions contained in the Code of Federal
Regulations (CFR) part 172.101 (49CFR Part 172).

The project was divided into three phases.

• The initial portion of Phase I characterized the shipment impacts for one year of Class 3 HM
  shipments and assessed the feasibility of conducting a comprehensive risk assessment of HM
  and non-HM shipments. Class 3 materials were selected because of their relative importance
  among HM shipments in volume and their potential for injury and damage during an accident.
  The characterization of the one-year of impacts of Class 3 HM shipments is contained in this
  report. The assessment of the feasibility of conducting a comprehensive risk assessment of HM
  and non-HM shipments is contained in the Plan for Assessing the Feasibility for Conducting a
  Comparative Risk Assessment on Hazardous Materials and Nonhazardous Materials
  Movements, April 1999.
  
  The second portion of Phase I characterized the shipment impacts for one year of Class 2.1 and
  Class 8 shipments as well as a preliminary annual portrait of non-HM shipments. These
  characterizations are also contained in this report.
  
  The project’s first phase also produced the two white papers: Potential for Integrating Hazmat
  Transportation Risk Assessment into SafeStat and The Identification of High Consequence Low
  Frequency Class 3 Hazmat Transportation Accidents. The papers were produced in late 1999.
  
• Phase II of the project produced the actual comparative risk assessment between HM and non-
  HM truck shipments. The overall HM risk depends on the risk associated with each class/
  division of hazardous material. They are calculated for this report. Thus, with the completion
  of Phase II of the project, the risk associated with the shipment of any class/division of HM can
  be compared to the risk associated with other classes/divisions as well as to the shipment of
  non-HM materials.
  
• Phase III of this project uses the information developed for the first two phases and analyzes the
  SafeStat algorithm to determine the appropriate inclusion of the risk of hazardous materials in
  the FMCSA carrier selection process.
  
  

1.2 Hazardous Materials Transportation

A hazardous material shipment is cargo that is part or all hazardous material according to the Code
of Federal Regulations (49CFR). An incident involving the shipment of HM is defined in 49 CFR
parts 171.15 and 171.16 and includes criteria for non-spill accidents. In the CFR, hazardous
materials are separated into the following classes (49CFR Part 171):

· Class 1 — Explosives
· Class 2 — Gases
· Class 3 — Flammable liquids (and combustible liquids)
· Class 4 — Flammable solids; spontaneously combustible materials and dangerous when wet
materials
· Class 5 — Oxidizers and organic peroxides
· Class 6 — Toxic (poison) materials and infectious substances
· Class 7 — Radioactive materials
· Class 8 — Corrosive materials
· Class 9 — Miscellaneous dangerous goods.

The majority of classes are segmented into divisions. For purposes of comparing risks, this analysis
employed a finer categorization of hazardous materials. Specifically, risks were developed for the
following classes and divisions or groups of divisions of HM. These are called categories in the
report.

· Class 1: Divisions 1.1, 1.2, 1.3 - all have the potential for mass detonating
· Class 1: Divisions 1.4, 1.5, 1.6 - characteristics make mass detonation extremely unlikely
· Class 2: Division 2.1 - Flammable Gases
· Class 2: Division 2.2 - Non-flammable Gases
· Class 2: Division 2.3 - Poisonous Gases
· Class 3
· Class 4: Division 4.1, 4.2, 4.3
· Class 5: Division 5.1, 5.2
· Class 6: Division 6.1, 6.2
· Class 7
· Class 8
· Class 9

This risk assessment considered a total of 12 different categories of hazardous materials. Adding
non-HM transport brings the total number of categories of materials assessed to 13.

1.3 Hazardous Material Flow

An essential element of the annual characterization of HM shipments is a description of traffic
flows. An estimate of transportation flows for all truck traffic and for all hazardous materials can
be derived from several sources.

One source is the 1993 Commodity Flow Survey (CFS) (U.S. Department of Commerce, 1996).
The (CFS) is a component of the quinquennial Census of Transportation that is designed to sample
the economic activity of the transportation of goods by mode of transportation. The 1993
Commodity Flow Survey provides an estimate of ton-miles for all commodities shipped and an
approximate estimate of the percentage of HM shipments of this total volume. The report shows
that all commodities were shipped an estimated 869,536,000,000 ton-miles in 1993 with hazardous
materials comprising about 74,410,000,000 ton miles of this total. Hazardous materials represent
about 8.5 percent of the total ton-miles. Unfortunately, the data for calculating the percentage of the
HM allocated to the various HM classes is limited, so the 1993 Commodity Flow study does not
provide a reasonable number in this regard. In addition, average shipment tonnages are not
available for calculating the mileage.

The 1997 CFS (U.S. Department of Commerce, 2000) is a more recent source of data. The report
shows that all commodities shipped by truck comprised an estimated 1,023,506,000,000 ton miles
in 1997, with hazardous materials comprising about 74,939,000,000 ton miles of this total. This
represents about 7 percent of the total truck ton mileage. Utilizing average tonnage values per
shipment supplied by the Census Bureau and assuming an average of about two shipments per
truckload, the ton mileage for all truck shipments in 1997 can be converted into an estimated
182,132,216,586 vehicle miles. HM shipments constitute approximately 7,763,282,762 vehicle
miles, or approximately 5 percent of the total mileage. The data clearly indicate that HM
shipments, although on average heavier than non-HM shipments, tend to travel shorter distances.
This is especially true for Class 3 shipments that involve gasoline and fuel oil.

Another source for vehicle miles traveled is the Federal Highway Administration’s (FHWA)
Highway Statistics for 1996, which provides annual vehicle miles for 1996. The total for all
combination and single unit trucks is 182,756,000,000 miles (U.S. DOT, 1997b). During the first
phase of the project, the study utilized the National Fleet Safety Survey for 1996 to estimate the
percentage of HM (Star Mountain Inc., 1997). For 1996, using a weighted average, 7.2 percent of
all trucks surveyed carried HM. To calculate the percentage of Class 3 materials carried by truck
for 1996, five regional HM commodity flow surveys were used. Based on the five surveys, the
project team estimated that 52 percent of HM vehicles carried flammable liquids. Appendix A
provides additional information from these flow studies.

The Research and Special Program Administration’s (RSPA) Office of Hazardous Materials Safety
in their 1998 study “Hazardous Materials Shipments” (US DOT, 1998) provided an estimate for
the number of daily shipments of hazardous materials and the number of tons shipped. This study,
based on a number of sources, estimates that all hazardous material truck shipments accounted for
about 769,000 shipments per day and about 1.4 billion tons shipped annually. Petroleum products,
which comprise the major part of the Class 3 shipments, accounted for an estimated 314,000
of these daily shipments and about 1.04 billion annual tons shipped. Chemical and allied products
accounted for about 445,000 daily shipments and “other” for about 10,000 daily shipments. The
RSPA study found that although only 43 percent of all HM tonnage is transported by truck, this
accounts for approximately 94 percent of all the individual shipments transported by truck.

2.0 Study Methodology

This section describes the methodology used for this report. Crucial portions of the
methodology include the
• review, selection and analysis of available data sources;
• estimation of the number of hazardous material and non hazardous material accidents and
incidents for the annual portrait; and
• measurement of impacts from these accidents and incidents.

2.1 Accident and Incident Data Sources

In this report, an incident is defined as an event involving the transportation of hazardous material
that results in an unanticipated cost to the shipper, carrier or any other party. An accident is an
incident that occurs when the vehicle transporting the goods is involved in a collision. The study
included HM accidents with a release, HM accidents with no release, loading/unloading with
release, and enroute leaks not caused by a vehicular accident. Non-spill accidents warranted
consideration in this study because severe consequences (e.g., injuries and fatalities) can still occur.
In addition, law enforcement and fire protection officials often treat any HM accident as a potential
spill even if no release of material is apparent.

An initial step in developing a risk assessment is estimating the number of accidents and incidents
reliably for a defined period of time. In the initial part of Phase I, an estimate of accidents and
incidents was developed for Class 3 truck shipments for the annual portrait. The estimate focused
on the Hazardous Materials Information System (HMIS) database and utilized several sources of
data to adjust the incidents and accidents reported in the HMIS. The adjustment was made in an
effort to reflect the actual number of incidents and accidents in a one-year period. During the
second part of Phase I, the methodology developed for Class 3 was applied to two additional
classes/divisions of HM: Division 2.1 - Flammable Gases and Class 8 - Corrosives.

Findings during Phase I affected the Phase II risk assessment work. Data analysis revealed that the
impacts from fires and explosions represented a series of impacts that should be separately assessed
whenever the data could support such a breakout. Another finding was the necessity of using more
than one year of accident data for the other 11 classes/divisions of HM, if similar accident statistics
were to be realized. For the analyses of Division 2.1 and Class 8 transport, initially three years of
data were used. Eventually, over nine years of data were used to obtain the statistics for all
12 categories of HM.

As the studies began to focus on the categories with less shipping exposure, some techniques
adjusting for underreporting had to change as well. Rather than look at several databases and
determine the amount of underreporting directly, the underreporting was estimated using factors
obtained from the detailed look at the first three categories of hazardous material, Classes 3 and 8
and Division 2.1. Even if time and money permitted using the accident reporting comparisons for
the other categories of hazardous material, it would have not been possible because only the HMIS
data covered the entire nine-year study period. For most of the other databases, only one or two
years of data were obtainable. The following sections describe the databases used in this effort.

Data identified and reviewed during initial research efforts associated with FMCSA’s hazardous
materials risk assessment study were from multiple sources and categories with varying detail.
Sources of data reviewed consisted of federal and state databases as well as research studies and
analytical reports. The categories reviewed were numerator data, characterized as hazardous
materials accidents/incidents or general commodity highway crashes, and denominator data,
consisting of the flow or movement of hazardous materials and general commodities.

The data assembled and reviewed may be categorized as generally being from a federal or state
database with input in some instances by local authorities or private companies. The federal
databases are collected and maintained by multiple administrations within the U.S. DOT as well as
the Commerce Department’s Census Bureau. These data are collected under different regulations,
utilizing disparate definitions under programs that have varying missions. The state databases have
issues of incompleteness and inconsistency primarily due to jurisdictional reporting variances
among the states as well as diversity in data processing capability. A review of the various
pertinent databases initially assembled for this project follows.

2.1.1 Federal Databases

Hazardous Materials Information System (HMIS). The HMIS is a system of databases maintained
and managed by the Office of Hazardous Materials Safety (OHMS) within the RSPA.
The major database in the HMIS and the most pertinent for the FMCSA risk assessment study is the
incident/accident database. This database dates back to 1971, contains more than 300,000 records,
and currently adds approximately 14,000 reports annually. Although the HMIS is a multi-modal
database, about 85 percent of the records are in the highway mode. The HMIS consists of incidents
where an unintentional release of a hazardous material in commerce occurs during the course of
transportation or is possibly imminent and results in the closure of a major artery or an evacuation
of the general public. Although the HMIS annually adds more than 10,000 truck transport-related
reports, an average of 250 reports represent highway accidents with the great majority
(approximately 200) involving cargo tanks.

Until recently, the intrastate carriers, those operating in only one state, were not required to file
HM incident reports. Thus, for most of the recording period, the HMIS reports encompassed motor
carriers that operate interstate and those that transport certain highly hazardous materials interstate.
This reporting requirement was extended to intrastate motor carriers on October 1, 1998. In
49CFR, Parts 171.15 and 171.16 provide the specific reporting requirements. As a result of the
distribution practices of some hazardous materials, such as gasoline, fuel oil, propane, and fertilizers
that are transported in large volumes by intrastate motor carriers, a substantial increase in HMIS
reports was predicted but has not been immediately realized. The HMIS is specifically designed to
capture information concerning the unintentional release of a hazardous material. Although an
accident checkbox is available on the HMIS report form, the only detailed information involving the
causation of an accident is found in the narrative section or in attachments.

For the purposes of FMCSA’s risk assessment study, the HMIS represents the only national
database of hazardous materials highway transportation accidents and incidents with details of the
material, packaging, and consequences involved. This database is mature, well maintained, and has
been extensively examined; as a result, its limitations can be identified. The consequences
associated with an incident are not comprehensive and in some instances the report form may not
even be complete. This deficiency, together with the lack of accident information, intrastate carrier
incidents and non-spill incidents, requires input from additional databases whose strengths will
complement the HMIS for conducting the risk assessment. Of all the databases, this database is one
of the more thoroughly checked and most inconsistencies have been eliminated. While it is clear
that some accidents must be reported, two carriers might experience essentially the same minor
incident and one will report it and the other will not. The minor incidents that are reported
dominate the truck transport records contained in the HMIS database.

Registration Database. The registration database for carriers, shippers, and offerors of certain
types or quantities of hazardous materials is contained within RSPA’s HMIS. An annual registration
form must be completed and submitted to RSPA that indicates the company’s primary activity
and the states in which the company operates. The registration database collects approximately
26,500 records annually and may be sorted by primary activity, whether the registrant is a carrier,
offeror or both and whether the registrant operates inter- or intrastate. Recent annual tabulations
show that of the 26,500 registrations received by RSPA, 2,820 are intrastate carriers and 731
indicate that they are both carriers and offerors on an intrastate basis. This database may prove
useful in estimating the lack of intrastate incidents not recorded in 1999.

News Clippings Database. The RSPA contracts with a private clipping service to provide
nationwide coverage of newspaper reports of hazardous material incidents. Copies of these
incidents are forwarded to RSPA for entry into an electronic database. This database supplements
HMIS data by compiling hazardous materials incidents not reported to RSPA. Paper copies of this
database were obtained from RSPA, and after review, data elements were entered into a separate
database for comparison with the HMIS database.

Safetynet MCMIS Database. The Motor Carrier Management Information System (MCMIS) is a
system of databases - not unlike RSPA’s HMIS - managed by the FMCSA. The Safetynet database,
also known as the accident file, is comprised of police accident reports (PAR) assembled by the
states and forwarded to the FMCSA. Each state has adopted the National Governors Association’s
(NGA) twenty-two uniform truck accident data elements on their PAR. This database was designed
to provide a census of truck accidents nationwide. Among the states, there is a wide variance
among the local jurisdictions that provide PARs for a state’s submittal into Safetynet. Because of
this wide diversity of reporting jurisdictions within the states, some states have a more comprehensive
data set in Safetynet than others. This database captures the general details of a crash, as well
as information on the vehicle and hazardous material cargo involved.

For the purposes of the FMCSA risk assessment study, Safetynet data files were requested for
eight selected states (PA, IN, IA, MN, CO, OR, OH, and CA). Six of these states belong to the
Performance and Registration Information Systems Management (PRISM) program that links
U.S. DOT’s information system to the states’ systems. The PRISM program began as a mandate
from Congress in the Intermodal Surface Transportation Efficiency Act (ISTEA) of 1991 to explore
the potential of linking the commercial vehicle registration process to motor vehicle safety. The
PRISM program includes two major processes: the Commercial Vehicle Registration Process and
the Motor Carrier Safety Improvement Process (MCSIP). These two processes work in parallel to
identify motor carriers and to hold them responsible for the safety of their operations. The six
states participating in the PRISM program are also part of an effort to improve the accuracy and
timeliness of data reported to the federal government. The two non-PRISM states selected, OH and
CA, were chosen because they produce additional state databases that were expected to be useful
for the purposes of the FMCSA risk assessment study. The Safetynet database proved very useful
in adjusting the HMIS database by adding intrastate carrier accidents and non-spill accidents.

Trucks in Fatal Accidents. The Trucks in Fatal Accidents (TIFA) is a database developed by the
University of Michigan Truck Research Institute (UMTRI) from the Fatal Accident Reporting
System (FARS) compiled by the U.S. DOT. Under contract to the U.S. DOT, UMTRI identifies
truck accidents in FARS and does extensive follow up on details of the fatal truck accident,
including the presence of hazardous materials as cargo. TIFA does not however contain any details
on the consequences resulting from a hazardous materials spill. This makes it difficult to compare
TIFA with other databases containing HM data.

Commodity Flow Survey (CFS). The processing of the 1997 CFS hazardous materials report was
completed in the spring of 2000. Much needed data regarding the flow of hazardous materials for
risk assessment studies is now available from the 1997 CFS. The largest contribution to hazardous
materials data comes from the highway data produced from the 1997 CFS. In sharp contrast to a
single HM table produced from the 1993 CFS, a total of 26 HM tables were produced from the 1997
CFS. All but four of the 1997 CFS HM tables had some application to hazardous material
transportation in the highway mode.

The 1997 CFS hazardous materials tables included fourteen tables concerned with HM class or
division, six mode specific tables, three state/geographic tables, and three tables on selected
materials. The tabulations of the 1997 CFS hazardous materials data was compiled using the
standard CFS breakout of tons, ton miles, average shipment distance and weight. These data were
tabulated utilizing the data set assembled in the 1997 CFS from responses containing a UN/NA
entry. Presentations of the 1997 CFS hazardous materials data were constructed from the UN/NA
data set.

In addition to the 1997 CFS hazardous materials tables, estimates were derived and employed in
this risk assessment study to establish the approximate number of miles hazardous materials were
transported by truck to help in the identification of the exposure level of hazardous materials on our
nations’ highways. The hazardous material tables from the 1997 CFS can be found in the document
1997 Commodity Flow Survey issued April 2000, EC97TCF-US (HM) RV. Included in this report
are twenty-six tables, an overview of the 1997 CFS, a review of the sample design, data collection,
and an estimation methodology and sample report forms and instructions.

Vehicle Inventory and Use Survey (VIUS). The Vehicle Inventory and Use Survey (VIUS), known
as the Truck Inventory Use Survey (TIUS) until 1992, is a component of the quinquennial Census
of Transportation and complements the CFS. The name change occurred because other vehicles
such as buses and recreational vehicles were intended to be added to the sample frame in 1997.
However vehicles other than trucks were not included in the sample so VIUS remains solely a truck
survey.

The 1997 VIUS was released in early 2000 and is now available for review and analysis. A hard
copy report has been published and the micro data is available on a CD ROM. The VIUS provides
figures for the number and type of trucks in operation, together with the physical and operating
characteristics of the country’s truck population. The format for hazardous materials data collection
in the VIUS involves an indication of whether the truck was used to transport placarded hazardous
materials, with a hazard class breakout. A broad breakout of the national percentage of trucks that
have carried hazardous materials by hazard class and equipment type is available. Limitations
associated with this database include definitional issues (e.g., a truck may also include a pickup, and
a placard must have been used) and little trailer information, as well as a limited sample of about
131,000 registered private and commercial trucks to draw on.

2.1.2 State Databases

State reports and databases were utilized for Ohio, California, and Colorado. They included reports
from the Public Utility Commission of Ohio and databases from the California Highway Patrol and
Colorado State Patrol. These databases focus on hazardous material incidents and provide an
independent source of data.

California Highway Patrol (CHP). The CHP maintains a database of all reported hazardous
material incidents. A subset of the CHP database was obtained from the CHP for analysis in
FMCSA’s risk assessment study. This database includes information on the actual incident,
hazardous material, and casualties but lacked carrier information and whether the incident was
actually an incident or accident. However, the database was able to provide enough information on
1996 Class 3 accidents to supplement the HMIS database.

Colorado State Patrol. The Colorado State Patrol also maintains a database of all reported
hazardous material incidents. The 1996 hazardous material incidents database was obtained for
analysis for Phase I of FMCSA’s risk assessment study. The database contains information
concerning the actual incident, along with detailed information on the hazardous material and
carrier information. Thus, the database was able to provide enough information to supplement
HMIS.

The Public Utilities Commission of Ohio (PUCO) Incident Reports. The PUCO provided copies
of HM incident reports from January 1, 1996 to mid 1998. These reports contained information on
the incident and carrier along with evacuation and road closure details. The reports were also very
valuable in that they typically contained a detailed description of the incident, an item missing in
most of the other databases. The PUCO reports were reviewed and data was extracted and entered
into a database for comparison to HMIS.

2.1.3 Other Databases

Dialogue (Newspaper Clippings). A search of newspaper clippings from the eight states was
completed to identify Classes 3, 2.1, and 8 accidents/incidents for the annual portrait. Those that
were identified were included in the adjustment of the HMIS database. Most of the articles also
provided additional detailed information about the accident/incident.

2.2.1 Selection of Reference Database

The first step was to select a reference database. For the purposes of OMC’s risk assessment study,
the HMIS represents the only national database of hazardous materials highway transportation
incidents with details of the material, packaging, and consequences involved, although these
consequences may not be comprehensive. The database is well maintained and carrier participation
is required. Deficiencies include a lack of accidents or incidents involving intrastate carriers
(although this deficiency is being corrected for FY 1999) and lack of coverage for no-spill HM
accidents. No-spill HM accidents should be included in an analysis because law enforcement and
fire protection officials often treat any HM accident as a potential spill even if no release of material
is apparent. Any accident involving a truck transporting HM should receive serious scrutiny from
officials and the DOT.

DOT has done an excellent job maintaining the integrity of the database as various changes have
been made in the definition of the classes/divisions of Hazardous Materials and as additional fields
have been added. For example, when the explosive categories were changed from A, B, and C to
numerical categories, the 1982 through 1990 records were modified to show the A through C class
accidents as 1.7 through 1.9. This enables a database search to go back as far as 1982 and get
meaningful accident data on the classes of HM.

When the project was started, 1996 was chosen as the base year for the analysis. At that time it
happened to be the last year for which complete data were available from all data sources. The first
analysis was for Class 3, flammable and combustible liquid transport. Because this single class
represents more than 50 percent of all HM truck transport, good statistics could be obtained by
looking at just one year. The first study during the second part of Phase I added two additional
classes/divisions of HM to the analysis, Division 2.1 (flammable gases) and Class 8 (corrosives).
Because these materials are involved in fewer accidents, the analysis base was expanded to 3 years
of records, 1995 - 1997. During Phase II, this analysis was subsequently extended to all classes/
divisions of HM truck transport. In the expanded categories of HM, some categories have few
incidents occurring in a given year. Thus, for the final analysis, data from 1990 through
March 1999 were used to create an annual portrait of HM impacts. This provided the greatest
quantity of HM incident data from which consequence and likelihood values could be obtained.

While data from 1982 on could have been used, prior to 1990, only total impact costs were
provided. Since breaking the total cost out into multiple cost categories is crucial for the risk
assessment, and pre-1990 data did not have this information, it was not used. While more than nine
years of data were used to evaluate consequences and likelihoods, the risk portrait continued to
describe one year. Whenever the approach is to collect data that covers several years in order to
consider the results to be representative of a year portrait, there is always a concern about trends.
Accident rate changes and cost escalation trends might be expected to be major concerns. However,
the data for the period 1982 to 1998 shows that the average total cost of an accident remained
constant. Furthermore, the total number of accidents reported each year did not seem to change
significantly over the 17-year period. While this result was somewhat surprising given the
significant cost increases in parameters such as the vehicle cost, the HMIS data provided no basis
for the selection of an escalation factor, so none was used. Although the costs were checked with
other sources to determine reasonableness, subsequent research and analyses should be conducted to
confirm whether increased accident costs occurred during these years.

2.2.2 Selection of Additional Databases

Additional databases with strengths complementing the HMIS for conducting the risk assessment
were consulted to supplement HMIS data with data on other spill accidents (especially intrastate
accidents) and non-spill accidents. In all cases, the additional databases covered fewer years.
However, because it was always possible to reduce the statistics to cover a single year, this
limitation was not significant. The greatest limitation was in the time period covered by the databases.
Where the databases covered the same time period, it was feasible to look at data from
a single year in a wide variety of databases and, in so doing, evaluate the underreporting that was
present in the databases. However, it was not feasible to continue to compare databases over many
years of data and for all HM classes. First of all, most databases, particularly the state databases,
are generated for a specific purpose and the information may only have been collected for a year or
two. Secondly, if accidents are very infrequent, which is the case for some of the classes/divisions
of HM, when one database misses one accident, the correction factor for underreporting of that
class/division of HM would be large and making such corrections would not be an accurate
representation of reality. Thus in the second part of Phase I, two additional classes/divisions —
Division 2.1, flammable gases and Class 8, corrosives — were used to compare multiple databases
for