FMCSA Safety Program Performance Measures Intervention Model: Roadside Inspection and Traffic Enforcement Effectiveness Assessment September 2002 (Updated) Prepared for: Federal Motor Carrier Safety Administration Office of Data Analysis and Information Systems Analysis Division, MC-RIA 400 Seventh Street, S.W. Washington, DC 20590 Prepared by: John A. Volpe National Transportation Systems Center Motor Carrier Safety Assessment Division, DTS-47 Kendall Square Cambridge, MA 02142 PREFACE This report documents the methodology and results from an improved model to measure the effectiveness of two of the key safety programs of the Federal Motor Carrier Safety Administration (FMCSA). The research was conducted by the Research and Special Programs Administration's (RSPA) John A. Volpe National Transportation Systems Center (the Volpe Center) in Cambridge, MA under a project plan agreement with the FMCSA. The work on FMCSA Program Performance Measures addresses the requirements of the Government Performance and Results Act (GPRA) of 1993, which obligates federal agencies to measure the effectiveness of their programs as part of the budget cycle process. Work on FMCSA Program Performance Measures was initiated during FY 93. In December 1994, a report titled "Office of Motor Carriers Safety Program - Performance Measurement" was prepared. That report provided a comprehensive breakdown of Office of Motor Carriers (OMC) safety programs and activities and described about a dozen potential evaluation models. (Note: The OMC later became the FMCSA.) Based on the OMC's review, the Volpe Center revised the report and recommended four evaluation models to assess the key OMC programs: roadside inspections conducted by participating states under the Motor Carrier Safety Assistance Program (MCSAP), on-site compliance reviews conducted by the OMC field offices and the states, commercial vehicle traffic enforcement also performed by the states under the MCSAP, and a comprehensive assessment of combined effects. Two initial evaluation models covering the roadside inspection program and the compliance review program were described in detail in a December 1998 report titled "OMC Safety Program Performance Measures." A review panel was convened to evaluate these models and made recommendations for improvement. The Volpe Center incorporated these recommendations together with other Volpe Center defined improvements into two "second-generation" models that measure the effectiveness of these two programs. This report describes the implementation of the Intervention Model, which covers not only the roadside inspection program, but also the traffic enforcement program. At the FMCSA, the project is managed by Dale Sienicki of the Office of Data Analysis and Information Systems, Analysis Division. The Volpe Center project manager is Donald Wright, Chief of the Motor Carrier Safety Assessment Division in the Office of System and Economic Assessment. The analysis was performed at the Volpe Center by Donald Wright, Dennis Piccolo and Emmett Harris of EG&G Services, under contract to the Volpe Center, with assistance from Dr. Thomas M. Corsi of the Supply Chain Management Center, Robert H. Smith School of Business, University of Maryland, College Park, Maryland. TABLE OF CONTENTS Section Page EXECUTIVE SUMMARY..................................................................................................................vii 1. INTRODUCTION......................................................................................................................... 1-1 1.1. Project Objective................................................................................................................ 1-1 1.2. Project Background............................................................................................................ 1-1 1.3. Project Scope...................................................................................................................... 1-1 1.4. Report Structure................................................................................................................. 1-2 2. SAFE-MILES: INITIAL MODEL............................................................................................... 2-1 2.1. Model Overview................................................................................................................ 2-1 2.1.1. Direct Effects....................................................................................................... 2-1 2.1.2. Indirect Effects..................................................................................................... 2-1 2.2. Model Limitations............................................................................................................... 2-2 3. INTERVENTION MODEL.......................................................................................................... 3-1 3.1. Model Description.............................................................................................................. 3-1 3.1.1. Crash Risk Probabilities...................................................................................... 3-2 3.1.2. Direct Effects....................................................................................................... 3-4 3.1.3. Indirect Effects..................................................................................................... 3-6 3.2. Implementation of the Intervention Model....................................................................... 3-7 3.3. Program Benefits............................................................................................................... 3-8 4. ENHANCEMENTS, APPLICATIONS, AND ANALYSES...................................................... 4-1 4.1. Introduction........................................................................................................................ 4-1 4.2. Intervention Model Enhancements.................................................................................... 4-1 4.2.1. Strengthen Crash Probabilities............................................................................ 4-1 4.2.2. Incorporate Hazardous Materials Violations...................................................... 4-2 TABLE OF CONTENTS (continued) Section Page 4.3. Intervention Model Applications....................................................................................... 4-2 4.3.1. Carrier Class Studies........................................................................................... 4-2 4.3.2. Alternate Treatments........................................................................................... 4-2 4.4. Future Intervention Model Analyses................................................................................. 4-2 4.5. Subsequent Model Runs……………................................................................................. 4-3 APPENDIX A. MATHEMATICAL DESCRIPTION OF THE INTERVENTION MODEL................................................................................................................. A-1 A.1. Overview.......................................................................................................................... A-1 A.2. Intervention Data.............................................................................................................. A-1 A.2.1. Roadside Inspections......................................................................................... A-1 A.2.2. Traffic Enforcements......................................................................................... A-2 A.3. Intervention-Level Impact................................................................................................ A-2 A.3.1. Violation Crash Risk Probability Profile.......................................................... A-2 A.3.1.1. Applied to Recorded Violations.................................................... A-5 A.3.1.2. Occurrences per Risk Category..................................................... A-5 A.3.2. Crashes Avoided per Intersection...................................................................... A-6 A.3.3. Examples............................................................................................................ A-7 A.4. Program-Level Impact..................................................................................................... A-9 A.4.1. Direct-Effect Approach..................................................................................... A-9 A.4.1.1. Primary Determination.................................................................. A-9 A.4.1.2. Roadside Allowance...................................................................... A-10 A.4.1.3. Examples...................................................................................... A-11 TABLE OF CONTENTS (continued) Section Page A.4.2. Indirect-Effect Approach................................................................................. A-12 A.4.2.1. Primary Determination................................................................. A-13 A.4.2.2. Roadside Allowance..................................................................... A-19 A.4.2.3. Examples...................................................................................... A-21 A.5. Program Benefits............................................................................................................ A-25 A.5.1. Fatal and Injury Crashes Avoided................................................................... A-27 A.5.2. Lives Saved...................................................................................................... A-29 A.5.3. Injuries Avoided............................................................................................... A-29 A.5.4. Examples.......................................................................................................... A-30 APPENDIX B. VIOLATIONS....................................................................................................... B-1 APPENDIX C. PROGRAM BENEFITS....................................................................................... C-1 C.1. National Program Benefits……………………………………………………………C-2 C.2. Roadside Inspection Benefits, by State………………………………………………..C-6 C.3. Traffic Enforcement Benefits, by State………………………………………………C-10 LIST OF ILLUSTRATIONS FIGURES Figure Page 3-1. Overview of Intervention Model............................................................................................... 3-1 3-2. Violation Crash Risk Probability Profile................................................................................... 3-2 3-3. Direct-Effect Approach.............................................................................................................. 3-5 3-4. Indirect-Effect Approach..................................................................................................... 3-6 3-5. Program Benefits Determination................................................................................................. 3-8 A-1. Direct-Effect Approach with Roadside Allowance................................................................. A-9 A-2. Indirect-Effect Approach with Roadside Allowance............................................................ A-13 A-3. Program Benefits Determination............................................................................................. A-26 TABLES Table Page 3-1. Relative Weights for Driver and Vehicle Violation Risk Categories....................................... 3-4 3-2. Data Inputs Used to Test the Model........................................................................................... 3-7 A-1a. Lower Bound Corrected Violation Estimates to Avoid One Crash, by Risk Category..... A-3 A-1b. Higher Bound Corrected Violation Estimates to Avoid One Crash, by Risk Category......……………………………………………………………………. A-4 A-2a. Lower Bound Crash Reduction Probabilities....................................................................… A-4 A-2b. Higher Bound Crash Reduction Probabilities.............................................................…….. A-4 A-3. Classifying Intervention Violations with the VRCCP: Two Examples.............................. A-5 A-4. Violation Occurrences per Risk Category: Two Examples.................................................. A-6 A-5. Indirect Effects Example Data.............................................................................................. A-21 B-1. Roadside Inspection Violations.............................................................................................. B-2 B-2. Traffic Enforcement Violations............................................................................................ B-14 C-1a. National Program Benefits, 1998........................................................................................... C-3 C-1b. National Program Benefits, 1999........................................................................................... C-4 C-1c. National Program Benefits, 2000........................................................................................... C-5 C-2a. Mean Roadside Inspection Program Benefits by State, 1998.............................................. C-7 C-2b. Mean Roadside Inspection Program Benefits by State, 1999.............................................. C-8 C-2c. Mean Roadside Inspection Program Benefits by State, 2000.............................................. C-9 C-3a. Mean Traffic Enforcement Program Benefits by State, 1998............................................. C-11 LIST OF ILLUSTRATIONS (continued) C-3b. Mean Traffic Enforcement Program Benefits by State, 1999............................................. C-12 C-3c. Mean Traffic Enforcement Program Benefits by State, 2000............................................. C-13 EXECUTIVE SUMMARY This report describes the Intervention Model, which is intended to provide the Federal Motor Carrier Safety Administration (FMCSA) with a means to gauge the effectiveness of two of its more critical safety programs – roadside inspections and traffic enforcements – in preventing crashes involving interstate motor carriers and in reducing related fatalities and injuries. The model is also intended to be a tool that the FMCSA can use periodically to measure the relative performance of its programs, and to analyze the effects of implementing different program changes. The model measures program effectiveness in terms of reductions in the numbers of crashes involving commercial vehicles, and in the numbers of associated fatalities and injuries. Although the methodology is believed to be sound and roadside inspection results are judged to be complete and accurate, the model suffers from several limitations resulting from a lack of empirical data regarding driver behavior and the contribution that vehicle defects and driver faults have on crash causation. Nevertheless, the model defaults to other means (including expert judgment) to compensate for these shortcomings and establishes a benchmark to measure roadside inspection and traffic enforcement program effectiveness. The model is based on the premise that the two programs – roadside inspection and traffic enforcement - directly and indirectly contribute to the reduction of crashes. As a result, the model includes two submodels that are used for measuring these different effects. Direct effects are based on the assumption that vehicle and/or driver defects discovered and then corrected as the results of interventions reduce the probability that these vehicles/drivers will be involved in subsequent crashes. The model calculates direct-effect-prevented crashes according to the number and type of violations detected and corrected during an intervention. Indirect effects are considered to be the by-products of the carriers' increased awareness of FMCSA programs and the potential consequences that these programs pose if steps are not taken to ensure and/or maintain higher levels of safety. In order to measure these indirect effects, which are essentially changes in behavior involving driver preparation and practices and vehicle maintenance, the model calculates responses to exposure to the programs and the resulting reduction in potentially crash-causing violations. Critical to the model is its ability to link vehicle and driver defects detected during inspections and/or traffic enforcement actions to crash probabilities. Currently available research and expert judgments provided the basis for establishing these linkages and assigning probabilities. Major investigations focusing on this linkage through special large truck crash data collections and crash reconstruction analysis are currently being sponsored by the FMCSA. The model's structure and analysis approach will enable the incorporation of the results of these efforts once they become available. The initial model run calculated the 1998 effects resulting from the roadside inspection and traffic enforcement programs. Subsequent model runs calculated program effects for 1999 and 2000. The table below displays the results. MCSAP Program Benefits: 1998-2000 1998 1999 2000 Roadside Inspections Crashes Avoided 8,612 9,119 9,362 Lives Saved 369 391 420 Injuries Avoided 5,902 6,250 6,416 Traffic Enforcement Crashes Avoided 2,800 3,021 3,306 Lives Saved 120 130 142 Injuries Avoided 1,919 2,071 2,265 This model, which measures the effectiveness of the roadside inspection and traffic enforcement programs, when combined with the Compliance Review Impact Assessment Model, forms a powerful performance measurement capability that will facilitate a combined-effects assessment of the three FMCSA safety programs. The expectation is that the combined-effects assessment results will further guide FMCSA decision-making when directing resources to achieve optimal program effectiveness. 1. INTRODUCTION 1.1. PROJECT OBJECTIVE The Intervention Model is designed to provide the Federal Motor Carrier Safety Administration (FMCSA) with a means to gauge the effectiveness of two of its more critical safety programs – roadside inspections and traffic enforcements – in preventing crashes involving interstate motor carriers and in reducing related fatalities and injuries. The model is also intended to be a tool that the FMCSA can use periodically to measure the relative performance of its programs, and to analyze the effects of implementing different program changes. Its use could provide a basis for making resource allocation and budgeting decisions that will help optimize the effectiveness and efficiency of the FMCSA's motor carrier safety programs. 1.2. PROJECT BACKGROUND During the 1980s, Congress passed several acts intended to strengthen motor carrier safety regulations. This led to the implementation of safety-oriented programs both at the federal and state levels, and an interest in establishing methods for measuring the effectiveness of these programs. The Surface Transportation Assistance Act of 1982 established the Motor Carrier Safety Assistance Program (MCSAP), a grants-in-aid program to states, to conduct roadside inspection and traffic enforcement programs aimed at commercial motor vehicles. The 1984 Motor Carrier Safety Act directed the U.S. Department of Transportation (U.S. DOT) to establish safety fitness standards for carriers. The U.S. DOT, along with the states, responded by implementing the MCSAP to fund roadside inspection and traffic enforcement programs, and the safety fitness determination process and rating system (based on on-site safety audits called compliance reviews). 1.3. PROJECT SCOPE The Program Performance Measures project established and managed by the FMCSA includes roadside inspection, traffic enforcement, and compliance review activities and programs. This report describes the development of a model, the Intervention Model, that is intended to measure the effectiveness of two of the three programs - roadside inspection and traffic enforcement - in reducing crashes and avoiding fatalities and injuries. It is believed that FMCSA safety program elements exert a positive influence, causing changes in driver behavior and carrier operations that lead to improvements in the level of motor carrier safety. At the same time, it is recognized that motor carriers are affected by exogenous influences, such as those attributable to the highway environment, that may intervene, impact or have some bearing on motor carrier safety. However, there is no accounting for these other influences and their associated consequences (i.e., fatalities and injuries) in this effort. Concurrent with the development of the Intervention Model, an improved model for measuring the effectiveness of compliance reviews (known as the Compliance Review Impact Assessment Model) was developed and documented. The ultimate plan is to assess the combined effects of all three programs. In the meantime, efforts to improve these safety program measures and models will continue independently, and the models will be run on a recurring basis to meet program objectives of measuring effectiveness, and to support annual budgetary planning and resource allocation decisions. 1.4. REPORT STRUCTURE This report includes descriptions of the evolution of the Intervention Model, the effects that it measures, and how the model is to be applied. The report also explains concepts driving the development process and affecting the model structure. Report sections include: ? Background on an earlier model, known as Safe-Miles, with an explanation of its limitations, ? A description of the model with results and descriptions of the calculation of direct and indirect effects, and ? A discussion of applications and future model enhancements. Technical appendices have been prepared that provide a mathematical description of the model (Appendix A), detailed information on the types and classification of violations critical to running the model (Appendix B), and program benefits as estimated by the model using MCSAP inspection/violation inputs (Appendix C). 2. SAFE-MILES: INITIAL MODEL 2.1. MODEL OVERVIEW The Safe-Miles Model that was also developed to measure the effectiveness of the roadside inspection program preceded the Intervention Model. It is discussed here by way of background, since the Intervention Model borrows substantially from the experience with the Safe-Miles Model. Included is a discussion of the direct and indirect effects approach first used in that model as well as the model's limitations leading to the development of the "second-generation" Intervention Model. The Safe-Miles Model employed a two-step analysis process to perform the evaluation. Instances were recorded in which vehicles and/or drivers were taken out of service during roadside inspections. Next, subsequent travel by the out-of-service (OOS) vehicles and drivers, once conditions were corrected, was converted into "safe miles" and estimates were made concerning crashes avoided during the "safe-miles" period. 2.1.1. Direct Effects Direct-effect benefits were accumulated from the point at which vehicles or drivers with OOS conditions were detected and removed from service. A three-month "safe" post-inspection period for vehicles was incorporated into the model. This time frame was considered appropriate since the Commercial Vehicle Safety Alliance (CVSA) has a three-month period after a vehicle receives a satisfactory inspection that it is exempt from additional inspections. Lacking an empirical basis with which to govern the duration of the direct effect findings for drivers, the post-inspection safe period for corrected driver OOS defects was shortened to a more conservative period of two months. 2.1.2. Indirect Effects Indirect effects are an equally important element of the roadside inspection program. The very existence of the program (as well as its magnitude) is believed to act as a deterrent. Knowledge of the program results in motor carrier managers making procedural changes that result in improvements in vehicle maintenance and inspection and in driver qualifications and behavior. These indirect effects, although assumed substantial, are much more difficult to quantify. The indirect effects are estimated in the Safe-Miles Model by assuming that carriers with a high frequency of (that is, greater exposure to) either vehicle or driver inspections, as a result of the enforcement of the roadside inspection program, change their behavior and voluntarily improve their safety, resulting in lower vehicle or driver OOS rates. Direct effects (crashes avoided) were added to indirect effects to derive total roadside inspection program benefits. These benefits were also expressed as estimates in dollar terms by using crash cost factors. There was no traffic enforcement component in the Safe-Miles Model. 2.2. MODEL LIMITATIONS The 1998 Volpe Center report - "OMC Safety Program Performance Measures" - identified the following limitations associated with the Safe-Miles Model: ? No observed evidence existed for the establishment of a driver safe-miles period. In future empirical studies of driver behavior, post-OOS violation detection would be required to establish the reliability of the two-month interval that was used. ? Each violation was considered in isolation. This precluded any heightening of the safety risk as a result of the presence of multiple violations found during an inspection. ? The lack of crash causation statistics hindered the ability to estimate the contribution of specified vehicle and driver defects to crash likelihood. The deterrence component of the model (indirect effects) relied on measured changes in OOS rates of carriers that had multiple inspections as a foundation for calculating indirect effects from roadside inspections. However, overall improved preparation and compliance of drivers and vehicles motivated by the presence of a roadside inspection program were thought to be greater than improvements that could be measured by the model. The research team defined the Intervention Model as a means to remedy these limitations. As with the Safe-Miles Model, the Intervention Model includes direct and indirect effect components; however, it: ? Eliminates the empirically weak "safe-miles" concept, ? Makes allowances for inspections with multiple violations, and ? Uses the latest available crash causation statistics to estimate the contribution of vehicle and driver faults to crash causation. The model also considers total inspection results. This means that it includes non-OOS violations, although with a lesser-assigned weight, in its calculations. Finally, the Intervention model remedies a Safe-Miles omission by including MCSAP program traffic enforcements in its analysis. The benefits of the Intervention Model are expressed as fatalities and injuries avoided as well as crashes avoided. 3. INTERVENTION MODEL 3.1. MODEL DESCRIPTION The Intervention Model was developed to determine the effectiveness of the MCSAP roadside inspection and traffic enforcement programs in reducing motor carrier crashes. The roadside inspection program consists of roadside inspections performed by qualified safety inspectors following the guidelines of the North American Standard, which was developed by the Commercial Vehicle Safety Alliance in cooperation with the FMCSA. Most roadside inspections by the states are conducted under a grant program (MCSAP) administered by the FMCSA. There are five levels of inspections including a vehicle component, a driver component or both. The traffic enforcement program is based on the enforcement of twenty-one moving violations noted in conjunction with a roadside inspection. Violations are included in the driver violation portion of the roadside inspection checklist. Figure 3-1 provides an overview of the Intervention Model. The diagram broadly illustrates: ? How the model begins with raw inspection violation data; ? Proceeds to the submodels, where separate algorithms are run to determine the direct and indirect effects; and ? Culminates, finally, with the calculation of program benefits for the respective programs. (For a mathematical description of the model, see Appendix A.) Figure 3-1. Overview of Intervention Model As with the Safe-Miles Model, this model is based on the premise that the two programs – roadside inspection and traffic enforcement - directly and indirectly contribute to the reduction of crashes. As a result, the model includes two submodels that are used for measuring these different effects. Direct effects are based on the assumption that vehicle and/or driver defects discovered and then corrected as the results of interventions reduce the probability that these vehicles/drivers will be involved in subsequent crashes. Indirect effects are considered to be the by-products of the carriers' increased awareness of FMCSA programs and the potential consequences that these programs pose if steps are not taken to ensure and/or maintain high levels of safety. 3.1.1. Crash Risk Probabilities In the model, the assumption is made that observed deficiencies (OOS and non-OOS violations) discovered at the time of roadside inspections and/or traffic enforcements can be converted into crash risk probabilities. This assumption is based on the premise that detected defects represent varying degrees of mechanical or judgmental faults, and, further, that some are more likely than others to play a contributory role in motor vehicle crashes. The assumption is that these deficiencies can be noted and ranked into discrete risk categories, each of which possesses a probability that reflects the crash risk that it poses. The process by which the resulting Violation Crash Risk Probability Profile (VCRRP) is formed appears in Figure 3-2. Figure 3-2. Violation Crash Risk Probability Profile The development of risk categories for violations relied upon a recent study conducted by Cycla Corporation. Each violation was classified according to the risk caused by the conditions of the violation. Cycla's report defined risk as "the likelihood of a violation leading to a crash" and, subsequently, divided the violations into five categories based on the level of risk. The risk categories and their descriptions are as follows: Risk Category 1 – The violation is the potential single, immediate factor leading to a crash or fatalities/injuries from a given crash. Risk Category 2 – The violation is the potential single, eventual factor leading to a crash or fatalities/injuries from a given crash. Risk Category 3 – The violation is a potential contributing factor leading to a crash or fatalities/injuries from a given crash. Risk Category 4 – The violation is an unlikely potential contributing factor leading to a crash or fatalities/injuries from a given crash Risk Category 5 – The violation has little or no connection to crashes or the prevention of fatalities/injuries. While covering most inspection violations, Cycla's assignment of violations to risk categories was incomplete. This required Volpe Center analysts to make violation assignments for those driver or vehicle violations not included in the Cycla risk assessment. These assignments were made based on comparability with the Cycla list. In the Cycla study, recommended weights were given to each of the risk categories, as shown in Table 3-1. The heaviest weight (1,000) was assigned to Risk Category 1 since these violations are considered to represent a significant safety hazard. Risk Categories 2 through 5 were given lesser weights (100, 10, 1, and 0.1, respectively). Cycla justifies this by stating that since "each relative numerical weight represents a different order of magnitude of likelihood, the weights decrease by a factor of ten." The Cycla study cautions, however, that the values do not refer to any "absolute" risk level. (The detailed list of roadside inspection violations and traffic enforcement violations, and associated risk categories appears in Tables B-1 and B-2 in Appendix B. Each table indicates the source of the categorization - either Cycla or Volpe Center.) To execute the model, Volpe Center analysts converted Cycla's relative numerical weights into crash reduction probabilities. Each probability is an estimate of the portion of a crash avoided when an inspection uncovers a particular violation. For example, if a violation carried a probability of 0.001, inspectors would have to discover that violation 1,000 times in order for the model to "take credit" for avoiding a crash. Since driver-related errors are thought to be more of a factor in crash causation relative to mechanical defects, traffic enforcement violations were assigned higher probabilities. In fact, a 4 to 1 ratio separates the two types of violations based on expert judgments formed from the results of previous studies and available data. Table 3-1. Relative Weights for Driver and Vehicle Violation Risk Categories Risk Category Relative Weight 1 Violation is the potential single, immediate factor leading to a crash or fatalities/injuries from a given crash. 1,000 2 Violation is the potential single, eventual factor leading to a crash or fatalities/injuries from a given crash. 100 3 Violation is a potential contributing factor leading to a crash or fatalities/injuries from a given crash. 10 4 Violation is an unlikely potential contributing factor leading to a crash or fatalities/injuries from a given crash. 1 5 Violation has little or no connection to crashes or the prevention of fatalities/injuries. 0.1 3.1.2. Direct Effects This section describes the methodology employed to estimate the number of direct-effect crashes avoided. Conceptually, the approach at the heart of the Direct Effects Submodel is straightforward. Since the occurrence of a single violation implies a certain degree of crash risk, each inspection that uncovers at least one violation can be interpreted as having reduced the risk linked with its noted violation(s). The model expresses this risk reduction in terms of the likelihood of a crash being avoided by each inspection violation that was noted and corrected. For an individual intervention, the avoided crash probability will be dependent upon the number and type of violations. Multiple violations, of course, will have a compounding effect, thereby increasing the likelihood of a prevented crash. By accounting separately for the two types of violations (roadside and traffic enforcement) and summing the portions of crashes avoided for all inspections within each group, it is possible to estimate direct-effect crashes that have been avoided due to the programs. Figure 3-3 depicts the process used to determine program direct effects. Figure 3-3. Direct-Effect Approach Four steps make-up the direct-effect approach. ? Step 1 - One year of inspection data is extracted from the Motor Carrier Management Information System (MCMIS) database. The MCMIS contains information compiled from federal and state safety agencies. Each intervention has its own set of associated driver and/or vehicle violations. ? Step 2 - An inspection's violations are matched to the Violation Crash Risk Probability Profile, whereby a list of crash reduction probabilities becomes attached to that inspection. This list becomes the basis for calculating the inspection's effect on avoiding a crash. ? Step 3 - The likelihood of an avoided crash for each inspection is calculated by using the crash reduction probabilities of the inspection. An inspection with multiple violations will have a greater likelihood of an avoided crash than will an inspection with a single violation. This result reflects the belief that multiple violations compound the safety hazard posed from driver deficiencies and/or vehicle defects. ? Step 4 - Once each inspection has been assigned its probability of avoiding a crash, the inspections are grouped by their initiating intervention. An inspection with a traffic enforcement driver violation is classified as traffic enforcement with a driver and/or vehicle roadside inspection component(s). All other inspections are classified as entirely driver and/or vehicle roadside inspections. Direct-effect crashes-avoided totals are simply the summation of 1) the portions of crashes avoided for all traffic enforcement violations and 2) the summation of the portions of crashes avoided for all roadside inspection violations. 3.1.3. Indirect Effects The fundamental premise of the indirect-effect approach is that once carriers have been exposed to the combination of roadside inspection and traffic enforcement actions, they will change their behavior. This change in behavior will result in higher levels of compliance, fewer future violations, and, therefore, a reduction in the number of crashes. This section presents a summary of the methods used in the model to arrive at program indirect effects. The deterrent-effects part of the model – that is, the Indirect Effects Submodel - follows a similar pattern to that of the Direct Effects Submodel. Indirect effects, by their nature, defy measurement. However, changes in behavior represented by changes in the number of violations recorded for a carrier over time can be used to identify and evaluate the results of the indirect effects. In other words, if a carrier receives fewer and fewer violations as it is subjected to more inspections, it will be determined that compliance behavior has been affected and the resulting likelihood of crashes has been reduced. To measure these effects, multiple successive years of intervention data are required. The Indirect Effects Submodel compares the results of inspections carrier by carrier from one year to the next in order to measure the effects of the exposure to having inspections on compliance. A carrier's performance in a base year is compared to its performance in a subsequent year. What is sought is an improvement, i.e., a reduction, in the likelihood of a crash resulting from increasingly fewer violations being recorded. The difference between the totals is calculated as the indirect-effect crashes-avoided effect. Depending upon the initiating intervention, it is tallied as indirect-effect crashes avoided for either the roadside inspection or traffic enforcement programs. Figure 3-4 illustrates the processes involved in assessing the indirect effects of the model. Figure 3-4. Indirect-Effect Approach The indirect effects calculation is similar to that of the direct effects. Steps 1 and 2 are equivalent, with one exception, to their counterparts in the Direct Effects Submodel. The Indirect Effects Submodel uses two years of MCMIS intervention data, whereas the Direct Effects Submodel uses one. Step 3 creates year one and year two average fractional crashes- avoided figures for each carrier. The two figures are compared and improvements are noted. Step 4 separates inspections and attributes the results to the initiating intervention. Traffic enforcement driver moving violations are assigned to the traffic enforcement program. All others (including driver and vehicle inspections done in conjunction with traffic stops) are assigned to the roadside inspection program. Indirect-effect crashes-avoided totals are the summation of the improvements in calculated crashes avoided. 3.2. IMPLEMENTATION OF THE INTERVENTION MODEL The use of the model requires intervention data inputs (as discussed in the submodel sections) in order to produce estimates of the numbers of crashes avoided that are attributable to the roadside inspection and traffic enforcement programs. For the purpose of testing the model, 1998 data was used, as shown in Table 3-2. Table 3-2. Data Inputs Used to Test the Model 1998 Total Interventions 2,217,000 Roadside Inspections with No Violations 572,000 Roadside Inspections with Violations 1,129,000 Traffic Enforcements with Violations 516,000 The Direct Effects Submodel yielded a mean estimate of 6,995 motor carrier crashes avoided as a result of the roadside inspection program in 1998, and another 2,331 crashes avoided due to the traffic enforcement program. The Indirect Effects Submodel, using the same 1998 input data, produced mean estimates of 1,617 roadside inspection and 469 traffic enforcement crashes avoided. Summation of the submodel totals provided estimates of the overall roadside inspection and traffic enforcement program results. Thus, the total numbers of crashes avoided in 1998 by the roadside inspection program and the traffic enforcement program were 8,612 and 2,800, respectively. 3.3. PROGRAM BENEFITS The model also estimates program benefits expressed in terms of lives saved and injuries avoided. Figure 3-5 illustrates the overall approach that is used by the model to determine these program benefits that are attributable to the roadside inspection and traffic enforcement programs. Figure 3-5. Program Benefits Determination Continuing with the 1998 data, the model converted the 8,612 crashes avoided by the roadside inspection program into program benefits of 369 lives saved and 5,902 injuries avoided. Likewise, the model converted the estimate of 2,800 crashes avoided as a result of the traffic enforcement into 120 lives saved and 1,919 injuries avoided. The set of tables in Section C.1 of the Appendix displays model-calculated national program results for calendar year 1998, as well as subsequent years for which the model was run. The model's flexibility lends itself to finer divisions of examination, such as scrutiny by state, which then can be used to guide the allocation of MCSAP resources and the design of state truck safety programs. The tables of Sections C.2 and C.3 in Appendix C show the estimated program benefits resulting from each state's MCSAP programs. Section C.2 tables show output from the model for state roadside inspections and the Section C.3 tables summarize traffic enforcement results. 4. ENHANCEMENTS, APPLICATIONS, AND ANALYSES 4.1. INTRODUCTION Additional model improvements are planned. They include improving the model inputs, such as the crash probabilities, and conducting additional assessments and analyses leading to improved application practices. Some of these improvements include: ? employing the results of planned studies of crash causation to improve crash probabilities, and capturing the compounding impact of multiple defects, ? incorporating hazardous materials violations, and the potential effect of these violations, particularly when combined with driver and vehicle effects, and ? determining the effectiveness of the programs in reducing crashes among different carrier classes allowing for an improved "targeting" of resources. Besides implementing model enhancements that will improve the measurement of the effectiveness of the roadside inspection and traffic enforcement programs, there will be ongoing efforts to examine how the model fits into a combined effects assessment of the three major FMCSA programs (including the compliance review (CR) program). Work will be initiated to establish an approach using the Intervention Model and the Compliance Review Impact Assessment Model to examine the combined effects and relative separate effectiveness of the programs. 4.2. INTERVENTION MODEL ENHANCEMENTS 4.2.1. Strengthen Crash Probabilities The Intervention Model is conservative in developing crash risk reduction probability estimates for individual violations as well as for individual inspections with multiple violations. Though the model clearly recognizes that multiple vehicle and driver problems occurring simultaneously greatly enhance the likelihood of a future crash, more empirical data on the compounding impact of multiple defects could result in much more accurate estimates of crash probabilities. While the Cycla effort to differentiate among violations based on their respective risk category provides a means to estimate the prospect that a crash would occur had the vehicle/driver not been stopped, further data on linkages between vehicle/driver problems and crash occurrences would improve the model's accuracy. The FMCSA and the National Highway Traffic Safety Administration (NHTSA) are currently conducting detailed post-crash investigations on a sample of crashes. The objective of this study is to obtain information on the connections between vehicle/driver problems and crash causation. 4.2.2. Incorporate Hazardous Materials Violations Another enhancement that will be made during future model runs is the effect of hazardous materials violations. Currently, the model does not address the issue of hazardous materials violations discovered during inspections or the effects that these violations (particularly when combined with driver and vehicle effects) may have on causing crashes or increasing the severity of crashes. This refinement is clearly warranted, given the potential effects of hazardous materials violations, especially when combined with vehicle and driver violations. 4.3. INTERVENTION MODEL APPLICATIONS 4.3.1. Carrier Class Studies By using motor carrier categories, or classes, such as those developed by Dr. Thomas Corsi of the Robert H. Smith School of Business at the University of Maryland, the model can be used to study program effectiveness among carrier classes. Differences in fleet size, driver age, length of haul, etc., may contribute to differences in direct-effect and indirect-effect program impacts. A better understanding of carrier classes and how they react to interventions will aid in the application and development of the roadside inspection and traffic enforcement programs. 4.3.2. Alternate Treatments As a corollary to the investigation of carrier types, alternate forms of treatment to reduce crashes should be sought. If patterns were to be discovered in particular strata of carriers, then the proposal and implementation of effective means of addressing these groups would become critical in the effort to increase the number of lives saved and injuries avoided from intervention programs. 4.4. FUTURE INTERVENTION MODEL ANALYSES The model is designed to be used as an ongoing measurement tool. It is anticipated that initial runs of the model will generate benchmarks that will assist in tracking program performance over time. In particular, emphasis should be placed on assessing the indirect effects component of the model, since it is the portion of the model that analyzes the effects that have an impact on future carrier behavior. Additional years of data would serve to substantiate the concept of the deterrence effect and improve the measurement of that effect as well. Finally, the results of the model are to be employed in a comprehensive assessment of the combined effects of all MCSAP safety programs. It is expected that combining the results of both the Compliance Review Impact Assessment and Intervention Models will create a more powerful program effectiveness measurement capability, which will enable the FMCSA to meet the requirements of the Government Performance and Results Act of 1993. The FMCSA will also employ this enhanced capability to improve the safety programs. 4.5. SUBSEQUENT MODEL RUNS Upon completion of the Model's initial testing, two subsequent Model runs were performed to determine program benefits in 1999 and 2000. These runs were to produce program performance benchmarks and act as a final test of the Model's ability to measure program performance across multiple years. Analysis of the new results showed an unanticipated drop in program benefits between 1998 and 1999. This occurred despite an increase in the overall number of interventions carried out in 1999. Further investigation of the underlying data was undertaken to discover the source of this apparent anomaly (e.g., an actual downward trend in program benefits, instability of the Model, etc.). Examination of the data uncovered a reporting inconsistency with one of the general driver violations (392.2D - "local laws/other driver violations"). Usage of this violation dropped dramatically between 1998 and 1999 and was offset by a concurrent increase in usage of another general driver violation (392.2 - "local laws/ general"). Whereas 392.2D was assigned to Risk Category 1, violation 392.2, as well as all other general driver violations, was classified as Risk Category 2. Consequently, Volpe Center staff opted to reclassify 392.2D as Risk Category 2 because a) it is a nonspecific violation that obscures the potential hazard of the behavior being cited, and b) doing so conforms with the classification established for other general driver violations. A new set of Model runs was completed after the reclassification for calendar years 1998, 1999, and 2000. The updated results appear in Appendix C. APPENDIX A. MATHEMATICAL DESCRIPTION OF THE INTERVENTION MODEL A.1. OVERVIEW The Intervention Model measures the effectiveness of the MCSAP roadside inspection and commercial vehicle traffic enforcement programs. Effectiveness, for the purposes of this analysis, is defined as the estimated reduction in motor carrier crashes attributable to the existence and implementation of the aforementioned MCSAP safety programs. The model is a key element of the FMCSA's Program Performance Measures project. This appendix presents a more detailed description of the model than that provided in the preceding text. It also contains mathematical explanations of the algorithms employed in the model. A.2. INTERVENTION DATA Raw intervention data serve as the inputs from which all further determinations flow. The data consist of individual records of roadside inspections and traffic enforcements carried out during a given period. The model creates a crashes-avoided figure for each intervention based on the number and type of violations present. A.2.1. Roadside Inspections Roadside inspections are interventions performed by qualified safety inspectors at fixed roadside locations (e.g., weigh stations) using North American Standard (NAS) guidelines. The NAS is a vehicle and driver inspection structure established by the FMCSA and the Commercial Vehicle Safety Alliance. A checklist of each roadside inspection lists uncovered violations of safety regulations. A.2.2. Traffic Enforcements MCSAP traffic enforcements are a subset of traffic enforcements in general. MCSAP traffic enforcements include only those enforcement stops that lead to an on-the-spot roadside inspection. The enforcement agent, if qualified, performs the subsequent roadside inspection. Otherwise, a safety inspector is called to the scene to conduct it. Since a traffic infraction precipitates the ensuing roadside inspection, 21 moving violations are incorporated into the driver section of the roadside checklist. The model classifies an intervention as traffic enforcement when at least one traffic violation is present in the intervention record. A.3. INTERVENTION-LEVEL IMPACT As the name implies, the Intervention Model places a great deal of importance on individual interventions. The reason for this is that violation tabulations come from interventions and those tabulations are matched against a Violation Crash Risk Probability Profile, which then serves as a basis for determining the number of crashes avoided for a given intervention. Aggregates developed from the intervention-level crashes avoided numbers eventually form national and state statistics. A.3.1. Violation Crash Risk Probability Profile The model assumes that observed deficiencies (OOS and non-OOS violations) can be converted into crash risk probabilities. This assumption is based on the belief that detected defects represent varying degrees of mechanical or judgmental faults and, as a result, some are more likely than others to play contributory roles in causing motor carrier crashes. These differences can be estimated and ranked into discrete risk categories. Thus, the Violation Crash Risk Probability Profile (VCRPP) contains all violation codes, each with an assigned risk category and a corresponding crash probability. Using Cycla's risk categories and the relative weights assigned to the categories, the Volpe Center analysts sought to account for error margins by opting for two probability sets – a Higher Bound set and a Lower Bound set. The outputs computed from the two sets are used to compute a mean with a range of ± 20 percent. Because crash causation data is still forthcoming, users are reminded to employ caution interpreting the Model's results. The figures in Tables A-1a and A-1b indicate the Higher Bound and Lower Bound numbers of violations that would have to be discovered to cause the model to credit one of the programs with an avoided crash. Keep in mind, however, the numbers in the tables are not meant to be definitive. They constitute the best guesses of industry experts interpreting available data. Volpe Center analysts used these figures to test and calibrate the model. As more reliable crash causation statistics become available, table quantities may have to be revised. These revisions will not affect the overall soundness of the model. Note that in moving from Risk Category (RC) 1 to RC 2, from RC 2 to RC 3, and so on, each step varies by a factor of ten. This tracks Cycla's variation in designated relative weights between risk categories. Note further that the weight given to uncovered traffic enforcement violations is four times that of the roadside inspection counterpart violations. Tables A-1a and A-1b illustrate the factor and weighting differences. For example, the tenfold factor variation can be seen when Traffic Enforcement RC1 OOS Violations jump from 30 to 300 when stepping to Traffic Enforcement OOS Violations RC2. Additionally, it takes quadruple the number of Roadside Inspection OOS Violations in RC1 (120) to have the same impact as Traffic Enforcement OOS Violations in RC1 (30), demonstrating the reduced weight given to roadside inspection violations vis-ΰ-vis traffic enforcement violations. Volpe Center analysts used the latest, preliminary data available from ongoing crash causation studies to support this difference. The studies found that driver faults represented by traffic enforcement violations are more likely to lead to motor carrier crashes than are roadside-inspection driver or vehicle faults of an equivalent risk category. Table A-1a. Lower Bound Corrected Violation Estimates to Avoid One Crash, by Risk Category Risk Category Roadside Inspection Traffic Enforcement Number of Violations Number of Violations OOS Violations Non-OOS Violations OOS Violations Non-OOS Violations 1 120 240 30 60 2 1,200 2,400 300 600 3 12,000 24,000 3,000 6,000 4 120,000 240,000 30,000 60,000 5 1,200,000 2,400,000 300,000 600,000 Table A-1b. Higher Bound Corrected Violation Estimates to Avoid One Crash, by Risk Category Risk Category Roadside Inspection Traffic Enforcement Number of Violations Number of Violations OOS Violations Non-OOS Violations OOS Violations Non-OOS Violations 1 80 160 20 40 2 800 1,600 200 400 3 8,000 16,000 2,000 4,000 4 80,000 160,000 20,000 40,000 5 800,000 1,600,000 200,000 400,000 Tables A-2a and A-2b display the higher bound and lower bound probabilities, respectively. The crash reduction probabilities are the reciprocals of the numbers in Tables A-1a and A-1b, so it follows that the probabilities also experience a tenfold change between steps. The crash reduction probabilities associated with each violation form the VCRPP. Table A-2a. Lower Bound Crash Reduction Probabilities Risk Category Roadside Inspection Traffic Enforcement Crash Reduction Probability Crash Reduction Probability OOS Violations Non-OOS Violations OOS Violations Non-OOS Violations 1 .00833 .004167 .033 .0167 2 .000833 .0004167 .0033 .00167 3 .0000833 .00004167 .00033 .000167 4 .00000833 .000004167 .000033 .0000167 5 .000000833 .0000004167 .0000033 .00000167 Table A-2b. Higher Bound Crash Reduction Probabilities Risk Category Roadside Inspection Traffic Enforcement Crash Reduction Probability Crash Reduction Probability OOS Violations Non-OOS Violations OOS Violations Non-OOS Violations 1 .0125 .00625 .05 .025 2 .00125 .000625 .005 .0025 3 .000125 .0000625 .0005 .00025 4 .0000125 .00000625 .00005 .000025 5 .00000125 .000000625 .000005 .0000025 A.3.1.1. Applied to Recorded Violations Because each inspection used in the analysis has one or more violations, the model classifies recorded violations according to their VCRPP ratings. Table A-3 displays the classification process for two example inspections. Inspection A is a roadside-initiated intervention, since no traffic enforcement violations are present. It contains roadside RC 1 OOS violations and both OOS and non-OOS RC 2 violations. Using the VCRPP, the violations receive their respective probabilities from the Higher Bound and Lower Bound probability sets. The VCRPP is also applied to Inspection B. Unlike Inspection A, Inspection B is classified as a traffic enforcement-initiated intervention, because it has at least one traffic enforcement violation. Additionally, several roadside violations were identified during the subsequent roadside inspection. Table A-3. Classifying Intervention Violations with the VCRPP: Two Examples A.3.1.2. Occurrences per Risk Category After the application of the VCRPP, the model aggregates violations occurring in a particular risk category. Table A-4 continues with the example interventions from Table A-3 by exhibiting the results of the aggregation. Table A-4. Violation Occurrences per Risk Category: Two Examples Inspection Roadside Inspection Traffic Enforcement Risk Category 1 Violations Risk Category 2 Violations Risk Category 1 Violations Risk Category 2 Violations OOS Non- OOS OOS Non- OOS OOS Non- OOS OOS Non- OOS A 2 3 4 B 2 2 1 2 A.3.2. Crashes Avoided per Intervention To generate an intervention's crashes avoided, the number of violation occurrences per risk category is multiplied by the crash probability associated with that risk category. For instance, if four occurrences of roadside OOS violations in RC 1 were noted on an inspection report, then the model would multiply four by the roadside OOS RC 1 probability from the VCRPP. This would be done for all roadside OOS and non-OOS violations, along with all traffic OOS and non-OOS violations. Summing the products creates an initial crash risk reduction for the inspection's risk category being evaluated. Initial Crash Risk Reduction per Risk Category (vrs-rcOOS X Prs-rcOOS) + (vrs-rcNON X Prs-rcNON) + (A-1) (vte-rcOOS X Pte-rcOOS) + (vte-rcNON X Pte-rcNON) = CRRrc-init where vrs-rcOOS = the number of roadside out-of-service violations in a given risk category recorded during an inspection, vrs-rcNON = the number of roadside non-out-of service violations in a given risk category recorded during an inspection, vte-rcOOS = the number of traffic out-of-service violations in a given risk category recorded during an inspection, vte-rcNON = the number of traffic non-out-of service violations in a given risk category recorded during an inspection, Prs-rcOOS = crash risk probability for a given roadside out-of-service risk category, Prs-rcNON = crash risk probability for a given roadside non-out-of-service risk category, Pte-rcOOS = crash risk probability for a given traffic out-of-service risk category, Pte-rcNON = crash risk probability for a given traffic non-out-of-service risk category, and CRRrc-init = initial, calculated crash risk for a given risk category within an inspection. Next, all violations recorded for a risk category during an intervention, roadside OOS and non- OOS and, if applicable, traffic OOS and non-OOS, are added together. Multiplying the total by the initial crash risk reduction calculated in Equation (A-1) produces the final crash risk reduction for a given risk category in a particular intervention. Equation (A-2) is designed to capture the growth in crash risk arising from the discovery and correction of numerous violations during a single intervention. The logic behind this is that, while each violation carries a certain degree of crash risk in isolation, additional violations occurring in tandem elevate the crash risk beyond the mere combined, additive, risk levels caused by each violation alone. In essence, the Final Crash Risk Reduction per Risk Category equation measures the multiplicative crash risk effect of compound safety defects. Final Crash Risk Reduction per Risk Category (vrs-rcOOS + vrs-rcNON + v te-rcOOS + v te-rcNON) X CRRrc-init = CRRRC (A-2) where CRRRC = final, calculated crash risk reduction for a given risk category within an inspection. Note: Equations (A-1) and (A-2) must be performed for each of the five risk categories. When all five risk categories have had their respective crash risk reductions determined, the model calculates the intervention's crashes avoided by adding the five CRRRC numbers. A cap of 0.75 is placed on the outcome for each intervention, thus ensuring that the model never produces a crashes avoided total greater than one. Volpe Center analysts chose three-quarters of a crash avoided as a cap to maintain a more conservative tendency in the model, given the lack of empirical crash causation data. Number of Crashes Avoided from an Intervention CRRRC1 + CRRRC2 + … + CRRRC5 = IA (A-3) where IA = calculated crashes avoided due to an inspection. Repeating this process using both Higher Bound and Lower Bound probabilities yields the crashes avoided range for each intervention. A.3.3. Examples Example A: In Inspection A (see Table A-3), a vehicle given a roadside inspection is found to have two out-of-service violations in Risk Category 1, three out-of-service violations in Risk Category 2, and four non-out-of-service violations in Risk Category 2. The calculation of the total crashes avoided of this single inspection, using Higher Bound probabilities, appears below. Multiplying the crash reduction probability for each risk category by the number of out-of- service violations in that risk category and adding it to the product of the risk reduction probability and the number of non-out-of-service violations gives the initial crash risk reduction. Thus, CRRrc-init for each risk category, based on Equation (A-1): Higher Bound Risk Category 1, CRRrc1-init (2 X .0125) = .025 Risk Category 2, CRRrc2-init (3 X .00125) + (4 X .000625)= .00625 Final crash risk reduction becomes known after multiplying the initial crash risk reduction for each risk category by the number of violations in that risk category. The model supplies total crashes avoided for the intervention by tallying the final crash risk reduction from each risk category. Inspection A's total crashes avoided, based on Equations (A-2) and (A-3): Higher Bound Risk Category 1, CRRRC1 .05 = .025 X 2 Risk Category 2, CRRRC2 + .04375 = .00625 X 7 Total Crash Risk Reduction, IA .09375 Therefore, Inspection A's range of crashes avoided begins at the Higher Bound result, 0.09375, and would extend to the Lower Bound output. Example B: In Inspection B (see Table A-3), a traffic enforcement stop has resulted in both traffic enforcement violations and roadside inspection violations. The intervention involved one traffic enforcement out-of-service violation in Risk Category 1 and two out-of-service violations in Risk Category 2. In addition, the inspection involved two roadside out-of-service violations in Risk Category 1 and two non out-of-service violations in Risk Category 2. Inspection B's computations follow: Higher Bound Roadside Traffic Risk Category 1, CRRrc1-init (2 X .0125) + (1 X .05) = .075 Using (A-1) Risk Category 2, CRRrc2-init (2 X .000625) + (2 X .005) = .01125 To account for multiple violations, the model makes the following intensification adjustments to calculate the final crash risk reduction for each risk category: Higher Bound Risk Category 1, CRRRC1 .225 = .075 X 3 Using (A-2) Risk Category 2, CRRRC2 + .045 = .01125 X 4 and (A-3) Total Crash Risk Reduction, IA .27 The crashes avoided range for Inspection B starts at 0.27. A.4. PROGRAM-LEVEL IMPACT Measuring interventions at the program level is next. It is here, however, that the model follows two divergent paths, one measuring direct effects and the other measuring indirect effects. Direct effects, it should be remembered, are the immediate products of roadside inspections and traffic enforcement stops performed in a given year, while indirect effects are based on behavioral changes caused by program awareness. A.4.1. Direct-Effect Approach This section outlines the development of direct-effect crashes-avoided estimates. Figure A-1 shows the process used to determine the direct effects of the programs. First, there is a primary crashes avoided computation. Afterwards, a roadside allocation credits a portion of traffic enforcement crashes avoided to the roadside inspection program, recognizing the contribution to the traffic total made by the ensuing roadside inspection. Figure A-1. Direct-Effect Approach with Roadside Allowance A.4.1.1. Primary Determination The model initially examines all inspections in a given year in terms of the numbers and types of violations associated with each individual inspection. Based on the VCRPP described above, inspection violations (both OOS and non-OOS) are matched with their respective crash risk reduction probabilities, to produce an estimated range of crashes avoided for that inspection. The model next segregates the complete set of inspections into two groups, depending on whether the initiating intervention was a roadside inspection or a traffic enforcement, and sums the estimated crashes-avoided ranges across all inspections in each group. Two overall estimates of crashes avoided emerge: one for the roadside inspection program and one for the traffic enforcement program. Roadside Inspection-initiated crashes avoided = IRS-A1 + IRS-A2 + … + IRS-An, (A-4) where IRS-A = crashes avoided per roadside inspection for (1, 2, …, n) roadside-initiated inspections. Likewise, Traffic Enforcement-initiated crashes avoided = ITE-A1 + ITE-A2 + … + ITE-Am, (A-5) where ITE-A = crashes avoided per traffic enforcement for (1, 2, …, m) traffic-initiated inspections. A.4.1.2. Roadside Allowance The process, however, does not end with the primary determination. An additional allocation of crashes avoided is necessary. As stated above, when the traffic enforcement action is the initiating event for an inspection, it is appropriate to credit back to the roadside inspection program those crashes avoided due to the correcting of roadside inspection-related violations. The model accomplishes the roadside allocation by using only the inspections initiated by traffic enforcement. Violations in this group are separated by type (roadside inspection-related and traffic enforcement-related) because two sets of crash risk reduction probabilities are required for each inspection. One set (A) is derived solely from traffic-related violations; the other (B) consists of the originally computed traffic enforcement crash risk reduction probabilities, using both types of violations. Dividing (A) by (B) provides the percentage of crashes avoided that need to be redistributed from the traffic enforcement program to the roadside inspection program. VTE Aadjustdirect = ---------- (A-6) VTE+RS where VTE = traffic enforcement-initiated crashes avoided from only traffic-related violations, VTE+RS = traffic enforcement-initiated crashes avoided from all violations, and Aadjustdirect = the percentage of traffic enforcement direct effect crashes avoided that will need to be allocated to the roadside inspection program. The final direct-effect program totals are then: RSA-direct = ARS-direct + [(1 - Aadjustdirect) X ATE-direct] (A-7) and TEA-direct = Aadjustdirect X ATE-direct (A-8) where ARS-direct = the pre-allocation crashes avoided total for roadside inspections, ATE-direct = the pre-allocation crashes avoided total for traffic enforcements, RSA-direct = the post-allocation direct effect crashes avoided total for roadside inspections, and TEA-direct = the post-allocation direct effect crashes avoided total for traffic enforcements. A.4.1.3. Examples Continuing with the example interventions, the results of applying Equations (A-5) through (A-8) to Inspection A and Inspection B appear below. Equation (A-5): Higher Bound Roadside Inspection-initiated crashes avoided = IRS-A1 = 0.09375 Traffic Enforcement-initiated crashes avoided = ITE-A1 = 0.27 Roadside Allowance, Equations (A-1), (A-2), (A-3): (Using Inspection B, the traffic enforcement-initiated intervention) Traffic Violations Only, Equation (A-1) Higher Bound Risk Category 1, CRRrc1-init (1 X .05) = .05 Risk Category 2, CRRrc2-init (2 X .005) = .01 Traffic Violations Only, Equation (A-2) Higher Bound Risk Category 1, CRRrc1-init .05 X 1 = .05 Risk Category 2, CRRrc2-init .01 X 2 = .02 Traffic Violations Only, Equation (A-3) Higher Bound Risk Category 1, CRRRC1 .05 Risk Category 2, CRRRC2 + .02 Total Crash Risk Reduction, IA .07 The crashes avoided range for Inspection B, using only traffic violations begins at 0.07. Applying Equation (A-6) gives the percentage of traffic enforcement-initiated crashes avoided that will be attributed to the traffic enforcement program. Higher Bound .07 ---------- = .259, i.e., 26% .27 Final direct effects crashes avoided, Equations (A-7) and (A-8). Roadside Total Traffic Total Higher Bound Higher Bound .09375 + [(1 - .26) X .27] = .29355 .26 X .27 = .0702 Thus, the recalculated higher bound crashes-avoided of the roadside program is 0.29, and the recalculated higher bound crashes-avoided of the traffic program is 0.07. A.4.2. Indirect-Effect Approach The fundamental premise of the indirect-effect approach is that once carriers have been exposed to the combination of roadside inspection and traffic enforcement actions, a change in their behavior will be manifested by a reduction in crashes. This section presents a summary of the methods used in the model to arrive at the programs' indirect effects. As with the direct-effect approach, a primary determination and a roadside allowance make up the major part of the procedure. Figure A-2 provides a view of the processes involved in assessing the indirect effects of the model. Figure A-2. Indirect-Effect Approach with Roadside Allowance Indirect effects require means other than direct measurement to reveal their presence. For that reason, the model uses changes in the number of violations recorded during inspections to identify and evaluate the indirect effects. Specifically, the model's algorithm employs two successive years of inspection data to undertake this process. To conduct a year-to-year comparison, it is necessary to identify and link the carriers who were inspected with the inspections each received during the two-year span. Only in this way can a cross-year evaluation discern the indirect influence (i.e., behavior modification) that causes a reduction in crashes. In contrast, this inspection-carrier link is not needed in the direct-effect approach. A.4.2.1. Primary Determination Gathered intervention data spanning two years is matched against the VCRPP, much in the manner laid out in the direct effects explanation. The model then organizes interventions by carriers. Intervention data from those carriers who have at least one intervention in both years are selected for preliminary analysis. The remaining Year One (Y1) intervention data, where a carrier match with Year Two (Y2) data was not able to be made, are set aside for later treatment. The nomenclature for the former group is Selected; the latter group is assigned the name Remaining. A.4.2.1.1. Selected Set For the Selected Set (S), the model determines each carrier's average crashes avoided in Y1, and again in Y2. The avoided crashes of each intervention from a given carrier in a given year (Equation (A-9)) are summed across the number of interventions the carrier had in that year (Equation (A-10)). Dividing the summation by all of the intervention actions conducted on the carrier for that year (Equation (A-11)) achieves the average crashes avoided. This provides Y1 and Y2 averages for each carrier in set S. Improved carriers in set S are those that have a decrease in average crashes avoided from Y1 to Y2. The improved subset designation applies to only those carriers with a lower Y2 figure. A crashes-avoided estimate for carriers in the improved subset of set S can now be made by multiplying the number of interventions a carrier had in Y1 by the difference in average crashes avoided it experienced between Y1 and Y2 (Equation (A-12) for roadside and Equation (A-13) for traffic). The model reaches the entire improved subset's crashes avoided aggregate by adding the crashes avoided totals for all of the carriers within the subset. A parallel summation for both the roadside inspection-initiated (Equation (A-14)) and traffic enforcement-initiated (Equation (A-15)) interventions supplies each program with a crashes avoided total from the improved subset of set S. Note: Calculate indirect effects separately, based on Higher Bound probabilities and Lower Bound probabilities. Carrier Crashes Avoided per Year Since every intervention has its own crashes avoided figure, summing the crashes avoided from each of the interventions a carrier received in a given year provides a crashes avoided total for that carrier. n ? IAh (A-9) h = 1 Carrier Interventions per Year Carrier interventions are the number of interventions a carrier had within a given year. n ? Ih (A-10) h = 1 where I = intervention, and IA = an intervention's crashes avoided for h (1, 2, …, n) interventions for a given carrier in a given year. Carrier Average Crashes Avoided per Year Using Equations (A-9) and (A-10), a carrier's average crashes avoided in a given year is calculated by dividing a carrier's crashes avoided by its total number of interventions. n ? IAh h = 1 CAavg = ----------- (A-11) n ? Ih h = 1 where CAavg = average crashes avoided for a given carrier in a given year. Carrier Crashes Avoided When a carrier's average crashes avoided diminishes in Y2, this is taken to be a positive indication of program indirect effects. Carriers who meet this condition are placed into an improved subset of set S called S?. Roadside The model determines an individual carrier's estimated number of roadside inspection crashes avoided resulting from indirect effects by taking the difference in its Y1 and Y2 average crashes avoided and multiplying the difference by the number of roadside inspections the carrier had in Y1. A modified version of Equation (A-10) that only counts roadside-initiated inspections from Y1 totals the number of roadside inspections. n ARS = (CAavg – Y1 - CAavg – Y2) X ? IY1-RS (A-12) RS = 1 where CAavg – Y1 = carrier average crashes avoided in Y1, CAavg – Y2 = carrier average crashes avoided in Y2, IY1-RS = Y1 roadside inspection, ARS = roadside inspection crashes avoided by a given carrier in subset S? due to RS (1, 2, …, n) roadside inspections in Y1, and the condition CAavg – Y1 > CAavg – Y2, or subset S?, is met. Traffic The model calculates traffic enforcement crashes avoided in a similar manner. n ATE = (CAavg – Y1 - CAavg) – Y2) X ? IY1-TE (A-13) TE = 1 where CAavg – Y1 = carrier average crashes avoided in Y1, CAavg – Y2 = carrier average crashes avoided in Y2, IY1-TE = Y1 traffic enforcement, ATE = traffic enforcement crashes avoided by a given carrier in subset S? due to TE (1, 2, …, n) traffic enforcements in Y1, and the condition CAavg – Y1 > CAavg – Y2, or subset S?, is met. Set S Preliminary Crashes Avoided Once Equations (A-12) and (A-13) have been used to create crashes avoided totals for each carrier in subset S?, preliminary program crashes avoided totals for set S are the aggregations of these totals. Roadside. m AS?-RS = ? ARSi (A-14) i = 1 where AS?-RS = set S roadside inspection crashes avoided for i (1,2, …, m) carriers in subset S?. Traffic m AS?-TE = ? ATEi (A-15) i = 1 where AS?-TE = set S traffic enforcement crashes avoided for i (1, 2, …, m) carriers in subset S?. A.4.2.1.2. Remaining Set Though crashes avoided have been calculated for the improved subset (S'), carrier and intervention data from the subset and its parent, set S, must still be used to impute crashes avoided totals to the Remaining Set (R). Because a definitive carrier-inspection link is absent over the course of Y1 and Y2, the R set requires estimations from general, intervention-related propositions. Therefore, two determinations are essential: the first is the ratio of interventions that are likely to be positively influenced by deterrence; the second characterizes the General Deterrence Impact of an intervention (described below). Since not all carriers in set S showed an improvement in their average crashes avoided from Y1 to Y2, the model assumes only a certain proportion of all interventions performed in Y1 carry an indirect influence. Dividing the total number of interventions in the improved subset (S') by the total number of interventions in the entire set S approximates the deterrence-to-intervention influence. The General Deterrence Impact (GDI) per intervention, on the other hand, attempts to quantify the portion of an avoided crash that is attributable to a single inspection, based again on the experience of the improved carrier subset. A unique GDI is calculated for each intervention type. The GDI for roadside inspections is the ratio of all improved subset roadside inspection crashes avoided divided by the total number of interventions in the subset, while the traffic enforcement GDI is the division of all improved subset traffic enforcement crashes avoided by the total number of interventions in the subset. Having determined these percentages, set R calculations may proceed. The percentage of interventions likely to be influenced by deterrence is multiplied by the total number of interventions in set R. The outcome is the estimated number of R interventions that would register an improvement in average crashes avoided. Next, the model estimates the number of indirect influenced set R interventions by the General Deterrence Impact per roadside inspection. The product of this calculation is the estimated roadside inspection crashes avoided for set R. Lastly, using the General Deterrence Impact per traffic enforcement, the same procedure develops R set estimated crashes avoided for traffic enforcement. The following equations, derived from Set S, provide the basis for estimating crashes avoided from Set R. Positive Influence of Deterrence. l ? Ij j = 1 D = ---------- (A-16) q ? Ik k = 1 where D = percentage of interventions positively influenced by deterrence, and I = inspection for j (1, 2, …, l) interventions in subset S? and for k (1, 2, …, q) interventions in set S. General Deterrence Impact Roadside The roadside inspection general deterrence impact is the ratio of all set S roadside inspection crashes avoided to the number of interventions (of either type) that are part of subset S?. AS?-RS GDIRS = ------------ (A-17) l ? Ij j = 1 where GDIRS = general deterrence impact per roadside inspection, and AS?-RS = set S roadside inspection crashes avoided for j (1, 2, …, l) interventions in subset S?. Traffic The traffic enforcement general deterrence impact is the ratio of all set S traffic enforcement crashes avoided to the number of interventions (of either type) that are part of subset S?. AS?-TE GDITE = ------------ (A-18) l ? Ij j = 1 where GDITE = general deterrence impact per traffic enforcement, and AS?-TE = set S traffic enforcement crashes avoided for j (1, 2, …, l) interventions in subset S?. Set R Indirect-Influenced Interventions With the results from Equation (A-16), it is possible to estimate the number of set R interventions that would be influenced by deterrence by multiplying the number of interventions in set R by the positive influence of deterrence. r RI = ? Ig X D (A-19) g = 1 where RI = the number of set R interventions positively influenced by deterrence, and D = the positive influence of deterrence for g (1, 2, …, r) interventions in set R. Set R Preliminary Crashes Avoided. Roadside The number of roadside inspection crashes avoided for set R is calculated by multiplying the general deterrence impact of a roadside inspection by the number of set R interventions positively influenced by deterrence. AR-RS = RI X GDIRS (A-20) where AR-RS = set R crashes avoided from roadside inspections, RI = the number of set R interventions positively influenced by deterrence, and GDIRS = general deterrence impact per roadside inspection. Traffic The number of traffic enforcement crashes avoided for set R is calculated by multiplying the general deterrence impact of a traffic enforcement by the number of set R interventions positively influenced by deterrence. AR-TE = RI X GDITE (A-21) where AR-TE = set R crashes avoided from traffic enforcements, RI = the number of set R interventions positively influenced by deterrence, and GDITE = general deterrence impact per traffic enforcement. A.4.2.2. Roadside Allowance Here too, the model allocates a portion of the crashes avoided derived from traffic enforcement actions back to the roadside program. Before doing so, overall indirect effect preliminary crashes avoided are obtained by adding the set S and R figures. Roadside ARS-indirect = AS?-RS + AR-RS (A-22) where ARS-indirect = the pre-allocation crashes avoided total for roadside inspections, AS?-RS = set S roadside inspection crashes avoided, and AR-RS = set R roadside inspection crashes avoided. Traffic ATE-indirect = AS?-TE + AR-TE (A-23) where ATE-indirect = the pre-allocation crashes avoided total for traffic enforcements, AS?-TE = set S traffic enforcement crashes avoided, and AR-TE = set R traffic enforcement crashes avoided. Equations (A-1), (A-2), and (A-3) are used to calculated crashes avoided totals for each intervention of the improved subset, using only the traffic-related violations. Dividing this by the results from Equation (A-15) provides the percentage of traffic enforcement-initiated crashes avoided that will need to be allocated to the roadside inspection program. V?TE Aadjustindirect = ---------- (A-24) V?TE+RS where V?TE = traffic enforcement-initiated crashes avoided from only traffic-related violations in subset S?, V?TE+RS = traffic enforcement-initiated crashes avoided from all violations in subset S?, and Aadjustindirect = the percentage of indirect effect traffic enforcement crashes avoided that will need to be allocated to the roadside inspection program. The final allocation of indirect effects is then: Indirect-effect crashes avoided from roadside inspections RSA-indirect = ARS-indirect + [(1 - Aadjustindirect) X ATE-indirect] (A-25) and Indirect effects crashes avoided from traffic enforcements TEA-indirect = Aadjustindirect X ATE-indirect (A-26) where ARS-indirect = the pre-allocation crashes avoided total for roadside inspections, ATE-indirect = the pre-allocation crashes avoided total for traffic enforcements, RSA-indirect = the post-allocation indirect effect crashes avoided total for roadside inspections, and TEA-indirect = the post-allocation indirect effect crashes avoided total for traffic enforcements. A.4.2.3. Examples Because indirect effects require more than a single year of data, the previous example interventions will not suffice. Therefore, a new set of example data appears in Table A-5. Y/C – Year/Carrier RC – Risk Category oos – out-of-service n-oos – non-out-of-service The first column identifies intervention data by carrier (Carrier A and Carrier B) over a two-year period. Note that Carrier A has interventions in both years, while Carrier B has interventions in Y1 only. This does not necessarily indicate that Carrier B had no interventions in Y2. Instead, it reflects the fact that interventions are not always able to be associated with a particular carrier and the model requires a carrier match in Y1 and Y2. Based on the criteria outlined in Section A.4.2.1, Carrier A would fall into the Selected Set and Carrier B would make up the Remaining Set. Equations (A-1) through (A-3) provide avoided crashes totals for each of the inspections in Table A-5. These figures form the input to the equations from the indirect-effect approach. Here, only the results created from the Higher Bound probabilities will be displayed. Lower Bound calculations follow the same steps. Summing the crashes avoided for each carrier in each year (Equation (A-9)) yields: Carrier A Crashes Avoided in Y1 0.05033 0.28487 0.00500 0.61023 + 0.05453 1.00496 Carrier A Crashes Avoided in Y2 0.00939 0.07092 0.02761 + 0.25032 0.35824 The number of interventions per carrier per year, Equation (A-10) Carrier A Number of Interventions in Y1 = 5 Carrier A Number of Interventions in Y2 = 4 Equation (A-11) supplies carrier average crashes avoided per carrier per year Carrier A Crashes Avoided in Y1 1.00496 --------------------------------------------------- = ----------- = 0.20099 Carrier A Number of Interventions in Y1 5 Carrier A Crashes Avoided in Y2 0.35824 --------------------------------------------------- = ----------- = 0.08956 Carrier A Number of Interventions in Y2 4 Carrier A's average crashes avoided in Y2 is less than the average in Y1. Thus, it meets the criterion to be included in the Improved Subset of the Selected Set. Indirect-effect roadside crashes avoided for Carrier A follow from Equation (A-12). (Carrier A Avg. Crashes Avoided in Y1 – Carrier A Avg. Crashes Avoided in Y2) X (Carrier A Number of Roadside Inspections in Y1) = (0.20099 – 0.08956) X 2 = 0.2229 Equation (A-13) supplies Carrier A's traffic crashes avoided. (Carrier A Avg. Crashes Avoided in Y1 – Carrier A Avg. Crashes Avoided in Y2) X (Carrier A Number of Traffic Enforcements in Y1) = (0.20099 – 0.08956) X 3 = 0.3343 The output of Equations (A-14) and (A-15) is, in this example case, identical to (A-12) and (A- 13), respectively, because Carrier A is the sole carrier within the Selected set. Were other carriers present, the outputs of (A-12) would be added to arrive at Selected set roadside inspection crashes avoided. Traffic enforcement crashes avoided would be the summation of the outputs from (A-13). Positive Influence of Deterrence, Equation (A-16) Number of Interventions in the Improved Subset 4 ------------------------------------------------------------ = ---- = 0.8 Number of Interventions in the Selected Set 5 General Deterrence Impact for Roadside Inspections, Equation (A-17) Selected Set Roadside Crashes Avoided 0.2229 ------------------------------------------------------------ = --------- = 0.04457 Number of Interventions in the Improved Subset 5 General Deterrence Impact for Traffic Enforcements, Equation (A-18) Selected Set Traffic Crashes Avoided 0.3343 ------------------------------------------------------------ = --------- = 0.06686 Number of Interventions in the Improved Subset 5 The calculations for the Remaining Set are next. Remaining Set Indirect-Influenced Interventions, Equation (A-19) = Number of Interventions in the Remaining Set X Positive Influence of Deterrence = 5 X 0.8 = 4 Remaining Set Preliminary Roadside Crashes Avoided, Equation (A-20) = Remaining Set Indirect-Influenced Interventions X General Deterrence Impact for Roadside Inspections = 4 X 0.04457 = 0.1783 Remaining Set Preliminary Traffic Crashes Avoided, Equation (A-21) = Remaining Set Indirect-Influenced Interventions X General Deterrence Impact for Traffic Enforcements = 4 X 0.06686 = 0.2674 Adding the Selected Set Crashes Avoided to the Remaining Set Crashes Avoided provides the pre-roadside allowance indirect-effects totals for each program. Roadside Inspection Preliminary Indirect Effect Crashes Avoided, Equation (A-22) 0.2229 + 0.1783 = 0.40115 Traffic Enforcement Preliminary Indirect Effect Crashes Avoided, Equation (A-23) 0.3343 + 0.2674 = 0.60173 Roadside Allowance, Equation (A-24) Traffic Enforcement Crashes Avoided from only Traffic-Related Violations --------------------------------------------------------------------------------------------- Traffic Enforcement Crashes Avoided from All Violations 0.13163 = ---------- 0.60173 = 0.22 Indirect Effects Crashes Avoided from Roadside Inspections, (A-25). Roadside Inspection Preliminary Indirect Effect Crashes Avoided + [(1 – Roadside Allowance) X Traffic Enforcement Preliminary Indirect Effect Crashes Avoided] = 0.40115 + [(1- 0.22) X 0.60173] = 0.8705 Indirect Effects Crashes Avoided from Traffic Enforcements, (A-26). Roadside Allowance X Traffic Enforcement Preliminary Indirect Effect Crashes Avoided = 0.22 X 0.60173 = 0.1324 A.5. PROGRAM BENEFITS Crash severity varies. Some crashes may result in no more than minor property damage, while others may result in bodily harm or loss of life. Of the many gradations possible, two classifications of crashes suffice for calculating program benefits, fatal crashes and injury crashes. Any motor carrier crash that results in at least one fatality is a fatal crash. A fatal crash may also involve injuries, but the fatality governs the crash's classification. Any motor carrier crash that results in at least one injury requiring transport for immediate medical attention but no fatalities, is an injury crash. Statistics of fatal and injury crashes supply the basis for creating lives saved and injuries avoided figures. This follows NHTSA established practice, which expresses program benefits in terms of lives saved and injuries avoided. Fatal crashes avoided translate to lives saved and injury crashes avoided translate to injuries avoided. Obtaining program benefits from estimated crashes-avoided figures requires two prior determinations, the first being a proportional identification of crashes by severity and the second being the average numbers of fatalities and injuries per crash. Fortunately, each has been completed elsewhere. According to a report done for the Federal Highway Administration's (FHWA) Office of Motor Carriers (OMC), of the trucks involved in crashes on U.S. roads in 1995, 3.6 percent were involved in fatal crashes, 40.0 percent were involved in injury crashes, and 56.4 percent were involved in towaway crashes. The average number of fatalities per fatal crash was calculated from data from the Fatality Analysis Reporting System (FARS), which is maintained by the NHTSA. For 1999 crashes involving large trucks or intercity buses, the ratio was 1.19 fatalities per fatal crash. The number of injuries per crash involves fatal as well as injury crashes, since fatal crashes can also result in injuries. State-reported crash data in the MCMIS were used to compute the average numbers of injuries in fatal and injury crashes. For 1999 large truck and bus crashes, the averages were as follows: ? Fatal crashes: 1.26 injuries per crash ? Injury crashes: 1.60 injuries per crash Figure A-3 shows the process used to calculate program benefits. Figure A-3. Program Benefits Determination Program Crashes Avoided (Direct and Indirect). The input to the program benefits portion of the model requires the union of crashes avoided attributable to direct effects and indirect effects. The program benefits calculations use the output of Equations (A-27) and (A-28). The calculations entail the development of estimated totals of crashes by severity as well as the final tally of lives saved and injuries avoided. Roadside RSA = RSA-direct + RSA-indirect (A-27) where RSA = roadside inspection crashes avoided from both direct and indirect effects, RSA-direct = the post-allocation direct-effect crashes avoided total for roadside inspections, and RSA-indirect = the post-allocation indirect-effect crashes avoided total for roadside inspections. Traffic TEA = TEA-direct + TEA-indirect (A-28) where TEA = traffic enforcement crashes avoided from both direct and indirect effects, TEA-direct = the post-allocation direct-effect crashes avoided total for traffic enforcements, and TEA-indirect = the post-allocation indirect-effect crashes avoided total for traffic enforcements. A.5.1. FATAL AND INJURY CRASHES AVOIDED The model breaks out program crashes-avoided figures into the numbers of program crashes avoided by severity. The proportions from the Center for National Truck Statistics report (9) mentioned previously are used by the model to calculate estimates of the numbers of fatal crashes and injury crashes avoided due to the roadside inspection and traffic enforcement programs. Roadside Multiplying the roadside crashes avoided from Equation (A-27) and the proportion of all highway crashes that resulted in fatalities provides the roadside fatal crashes avoided. Roadside injury crashes avoided are calculated similarly, only substituting the injury proportion of all highway crashes in place of the fatality proportion. RSA-Fatal = RSA X CSPFatal (A-29) RSA-Injury = RSA X CSPInjury (A-30) where RSA-Fatal = number of fatal crashes avoided due to the roadside inspection program, RSA-Injury = number of injury crashes avoided due to the roadside inspection program, RSA = number of roadside inspection crashes avoided, CSPFatal = proportion of all crash types that are fatal crashes, and CSPInjury = proportion of all crash types that are injury crashes. Traffic Fatal crashes for the traffic enforcement flow from Equation (A-28). TEA-Fatal = TEA X CSPFatal (A-31) TEA-Injury = TEA X CSPInjury (A-32) where TEA-Fatal = number of fatal crashes avoided due to the traffic enforcement program, TEA-Injury = number of injury crashes avoided due to the traffic enforcement program, TEA = number of traffic enforcement crashes avoided, CSPFatal = proportion of all crash types that are fatal crashes, and CSPInjury = proportion of all crash types that are injury crashes. A.5.2. LIVES SAVED To calculate the number of lives saved, the number of fatal crashes avoided is multiplied by the average number of fatalities per fatal crash. Roadside LSRS = RSA-Fatal X FCFatal (A-33) where LSRS = lives saved due to the roadside inspection program, RSA-Fatal = number of fatal crashes avoided due to the roadside inspection program, and FCFatal = average fatalities per fatal crash. Traffic LSTE = TEA-Fatal X FCFatal (A-34) where, LSTE = lives saved due to the traffic enforcement program, and TEA-Fatal = number of fatal crashes avoided due to the traffic enforcement program, and FCFatal = average fatalities per fatal crash. A.5.3. INJURIES AVOIDED To calculate the number of injuries avoided, the number of fatal crashes avoided is multiplied by the average number of injuries per fatal crash, and the number of injury crashes avoided is multiplied by the average number of injuries per injury crash. The two products are then added to obtain the total number of injuries avoided. Roadside IARS = (RSA-Fatal X FCInjury) + (RSA-Injury X ICInjury) (A-35) where IARS = number of injuries avoided due to roadside inspections, RSA-Fatal = number of fatal crashes avoided due to the roadside inspection program, RSA-Injury = number of injury crashes avoided due to the roadside inspection program, FCInjury = average injuries per fatal crash, and ICInjury = average injuries per fatal crash. Traffic IATE = (TEA-Fatal X FCInjury) + (TEA-Injury X ICInjury) (A-36) where TEA-Fatal = number of fatal crashes avoided due to the traffic enforcement program, TEA-Injury = number of injury crashes avoided due to the traffic enforcement program, FCInjury = average injuries per fatal crash, and ICInjury = average injuries per fatal crash. A.5.4. EXAMPLES Program Crashes Avoided Roadside Program Crashes Avoided (Direct and Indirect), (A-27) = Roadside Program Direct-Effect Crashes Avoided + Roadside Program Indirect-Effect Crashes Avoided = 0.9355 + 0.8705 = 1.806 Traffic Program Crashes Avoided (Direct and Indirect), (A-28) = Traffic Program Direct-Effect Crashes Avoided + Traffic Program Indirect-Effect Crashes Avoided = 0.0702 + 0.1324 = 0.203 Fatal Crashes Avoided Roadside Fatal Crashes Avoided, (A-29) = Roadside Program Crashes Avoided (Direct and Indirect) X Fatal proportion of truck crashes = 1.806 X 0.036 = 0.065 Traffic Fatal Crashes Avoided, (A-31) = Traffic Program Crashes Avoided (Direct and Indirect) X Fatal proportion of truck crashes = 0.203 X 0.036 = .0073 Injury Crashes Avoided Roadside Injury Crashes Avoided, (A-30) = Roadside Program Crashes Avoided (Direct and Indirect) X Injury proportion of truck crashes = 1.806 X 0.400 = 0.7224 Traffic Injury Crashes Avoided, (A-32) = Traffic Program Crashes Avoided (Direct and Indirect) X Injury proportion of truck crashes = 0.203 X 0.400 = 0.0812 Lives Saved Roadside Lives Saved, (A-33) = Roadside Fatal Crashes Avoided X Average fatalities per fatal crash = 0.065 X 1.19 = 0.0774 Traffic Lives Saved, (A-34) = Traffic Fatal Crashes Avoided X Average fatalities per fatal crash = 0.0073 X 1.19 = 0.0087 Injuries Avoided Roadside Injuries Avoided, (A-35) = (Roadside Fatal Crashes Avoided X Average fatalities per injury crash) + (Roadside Injury Crashes Avoided X Average. injuries per injury crash) = (0.065 X 1.26) + (0.7224 X 1.60) = 1.2377 Traffic Injuries Avoided, (A-36) = (Traffic Fatal Crashes Avoided X Average. fatalities per injury crash) + (Traffic Injury Crashes Avoided X Average injuries per injury crash) = (0.0073 X 1.26) + (0.0812 X 1.60) = 0.1391 APPENDIX B – VIOLATIONS Table B-1. Roadside Inspection Violations Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-1. Roadside Inspection Violations (continued) Table B-2. Traffic Enforcement Violations Table B-2. Traffic Enforcement Violations (continued) APPENDIX C – PROGRAM BENEFITS C.1. NATIONAL PROGRAM BENEFITS Table C-1a. National Program Benefits, 1998 Table C-1b. National Program Benefits, 1999 Table C-1c. National Program Benefits, 2000 C.2. ROADSIDE INSPECTION BENEFITS, BY STATE Table C-2a. Mean Roadside Inspection Program Benefits by State, 1998 Table C-2b. Mean Roadside Inspection Program Benefits by State, 1999 Table C-2c. Mean Roadside Inspection Program Benefits by State, 2000 C.3. TRAFFIC ENFORCEMENT BENEFITS, BY STATE Table C-3a. Mean Traffic Enforcement Program Benefits by State, 1998 Table C-3b. Mean Traffic Enforcement Program Benefits by State, 1999 Table C-3c. Mean Traffic Enforcement Program Benefits by State, 2000 Mean estimates. Higher and lower bound estimates were based on different risk assumptions, which may be found in Intervention Model: Roadside Inspection and Traffic Enforcement Effectiveness Assessment, Sept. 2002. Revised figures. See Section 4.5 for details. Except under the following circumstances: 1) A North American Commercial Vehicle Critical Safety Item or OOS violation is detected, 2) When a Level IV (Special Inspection) exercise is involved, 3) When a statistically-based random inspection technique is being employed to validate an individual jurisdiction or regional OOS percentage, or 4) When inspections are conducted to maintain CVSA inspection quality assurance. Commercial Vehicle Safety Alliance website, http://www.cvsa.org/Inspections/CVSA_Decals/cvsa_decals.html, 2001. For a complete list of driver and vehicle violations associated with the roadside inspections and traffic enforcement, see Appendix B. Cycla Corporation, Risk-based Evaluation of Commercial Motor Vehicle Roadside Violations: Process and Results, July 1998. Note: The twenty-one traffic enforcement violations that fall under MCSAP were also included in the Cycla evaluation. See Appendix A for the explanation of how the relative weights from Cycla were converted into crash risk probabilities. Based on preliminary findings from crash causation studies conducted by the University of Michigan Transportation Research Institute. An ongoing, more comprehensive crash causation study at the NHTSA is expected to bolster these assumptions. Ibid, p. 21. Readers should note that the allocation of violations to programs actually occurs earlier in the indirect-effect calculation process. To simplify the presentation, however, the submodel has been presented in the form appearing above. This does not materially affect the model outline. To determine indirect effects, the Model looked at carriers that had interventions in 1998 and 1999, then noted the difference between the two years' data. This was done because behavioral changes (i.e., indirect effects) brought about by 1998 interventions will only be seen through the impact that they have upon a carrier/driver over the course of the following year. Source: MCMIS file, March 2001. Figures appearing in the table have been rounded to the nearest thousand. Model output figures represent the mean between totals derived from two sets of crash risk probabilities. An explanation of the probability range and its effects on the model appears in Appendix A. The U.S. Department of Transportation's Federal Motor Carrier Safety Administration (FMCSA) and National Highway Traffic Safety Administration (NHTSA) are conducting the Large Truck Crash Causation Study. The Motor Carrier Safety Improvement Act of 1999 (MCSIA) provided for the study. "The MCSAP is a Federal grant program that provides financial assistance to States to reduce the number and severity of accidents … involving commercial motor vehicles (CMVs). … Investing grant monies in appropriate safety programs will increase the likelihood that safety defects, driver deficiencies, and unsafe motor carrier practices will be detected and corrected before they become contributing factors to accidents." http://www.fmcsa.dot.gov/safetyprogs/mcsap.htm. See http://www.inspector.org/37stepin.htm. § Sec.350.111 of the Federal Motor Carrier Safety Regulations defines a MCSAP traffic enforcement as follows: "Traffic enforcement means enforcement activities of State or local officials, including stopping CMVs operating on highways, streets, or roads for violations of State or local motor vehicle or traffic laws (e.g., speeding, following too closely, reckless driving, improper lane change). To be eligible for funding through the grant, traffic enforcement must include an appropriate North American Standard Inspection of the CMV or driver or both prior to releasing the driver or CMV for resumption of operations." Crash causation studies are underway at the University of Michigan Transportation Research Institute and the NHTSA. Ibid. To avoid needless complexity, the examples have been crafted using risk categories 1 and 2, rather than the entire range of risk categories 1 through 5. Note: Since only two example interventions have been presented, one roadside-initiated (Inspection A) and the other traffic-initiated (Inspection B), Equation (A-5)'s example results are identical to the output of Equation (A-3). An area for future investigation consists of motor carriers who registered no improvement in average crashes avoided. Center for National Truck Statistics, University of Michigan Transportation Research Institute, Truck and Bus Crash Factbook 1995, 1997. The Federal Highway Administration's (FHWA) Office of Motor Carriers (OMC) later became the Federal Motor Carrier Safety Administration (FMCSA). A towaway crash results in no fatalities or injuries requiring transport for immediate medical attention, but in one or more motor vehicles incurring disabling damage as a result of the crash, requiring the vehicle(s) to be transported away from the scene by a tow truck or other motor vehicle. 8 viii 1-2 2-2 3-8 4-3 A-32 B-15 C-13