Estimating the Presence of Alcohol and Drug Impairment in Traffic Crashes and their Costs to Canadians: A Discussion Paper
Submitted to MADD Canada by:
Applied Research and Evaluation Services (ARES)
2125 Main Mall
University of British Columbia
Vancouver BC
V6T 1Z4
December 2002
All information, opinions, and conclusions found in this report are the result of research by the Applied Research and Evaluation unit at the University of British Columbia (ARES) and do not necessarily represent the positions of any other individuals, agencies or groups. Copyright ARES, 2002.
Dr. Michael Marshall and Dr. Bill Mercer of ARES would like to thank the Traffic Injury Research Foundation (TIRF) and the Insurance Corporation of British Columbia (ICBC) for their assistance in the production of this report.
1
Table of Contents
Table of Contents
Table of Tables
Executive Summary
Argument:
Calculation of Frequencies:
Calculation of Costs:
Conclusion:
Introduction
An Overview of Some Common Crash Data Sources and their Limitations
Crash Severity & Frequency Data Sources
Police Data:
Insurance Data:
Coroner Data:
Alcohol and Drug Involvement Data Sources
Police Data
Coroner Data
Filling in the Gaps
Crash and Victim Counts
Fatal Crashes and Fatally injured Victims
Blood Alcohol Content (BAC) Count Estimations:
Alcohol-only, Alcohol-and-Drug, and Drug-only Impairment Estimations
An Order-of-Magnitude Calculation
Facts and Arguments
Assumptions:
Estimation:
Three Crash Costing Models
Three Perspectives:
Real Dollar Estimates:
Discounted Future Earnings:
Willingness to Pay:
Impaired Driving Crash Cost Estimations
Conclusion
References
Appendix A: Insurance Corporation of British Columbia and Manitoba Public Insurance Data & Transformations
ICBC
MPI
Table of Tables
Estimated Fatalities, Injuries & PDO Vehicles, Canada, 1999
Estimated % Alcohol-Involved - Fatalities, Injuries & PDO Vehicles, Canada, 1999
Estimated Impairment Source - Fatalities, Injuries & PDO Vehicles, Canada, 1999
Estimated # Impaired-related Fatalities, Injuries & PDO Vehicles, Canada, 1999
Crash costs by Costing Model in 1999$
Estimated Cost of Impaired Driving Crashes by Costing Models
Table 1: Motor Vehicle-Related Fatalities, BC, 1998
Table 2: # Motor Vehicle Crash Types, BC & Manitoba, 1998
Table 3: Injury Severity - Miller & Blincoe 1994
Table 4: ICBC Victim & PDO Vehicle to Fatality Ratios
Table 5: MPI Victim & PDO Vehicle to Fatality Ratios
Table 6: Injury Severity & BAC - Miller & Blincoe
Table 7: Estimated Fatalities, Injuries & PDO Vehicles, Canada, 1999
Table 8: % Impairment Source by Crash Type
Table 9: # Victim/ PDO Vehicle - Impairment Source by Crash Type
Table 10: Crash costs by Costing Model in 1999$
Table 11: Estimated Cost of Impaired Driving Crashes by Costing Models
Table 12: Impairment by Alcohol only Crash Costs by Costing Model
Table 13: Impairment by Alcohol & Drugs Crash Costs by Costing Model
Table 14: Impairment by Drugs only Crash Costs by Costing Model
Table A1: ICBC Crash Types N, 1996 - 2001
Table A2: ICBC Crash Fatality Ratios
Table A3: ICBC N victims/vehicles
Table A4: ICBC Victim & PDO Vehicle to Fatality Ratios
Table A5: MPI N victims/vehicles
Table A6: MPI Victim & PDO Vehicle to Fatality Ratios
1
Executive Summary
Knowledge of the extent of harm caused by traffic crashes, and by the sub-set of crashes caused by impairment, is important in the development of public policy and the allocation of countermeasure resources.
This discussion paper argues that the extent of injury and property damage only (PDO) crashes is generally underestimated and hence the magnitude of the subset of impaired crashes within those categories is also underestimated. It also argues that the extent of drug-impaired crashes is underestimated. A model is presented that attempts to address these sources of underestimation by working from essentially complete data on fatalities and projecting estimations to injury and PDO events. Cost estimations are then calculated using three different costing models.
Argument:
The more serious a crash, the more likely it will be reported to or otherwise become known to various authorities such as the police, motor vehicle branches, insurance companies, and the coroner, and the more likely it will be investigated by one or more of those bodies. Consequently, Canadian data on motor vehicle fatalities, and whether or not the fatally injured person(s) had a measurable blood alcohol content (BAC), is largely complete, valid and reliable. That is, we have very good information the number of persons killed in crashes, and whether or not they were possibly impaired by alcohol. On the other hand, the data on whether or not a fatally injured person might have been impaired by drugs is incomplete, primarily due to a lack of testing and testing sensitivity.
As crashes become less serious, there is less likelihood that they will be reported, recorded, or investigated. In order to assess the magnitude of the traffic crash and impairment-caused traffic crash problem, there is a need to find a way to estimate the number of less severe crashes, and whether or not they might have been caused by impairment by alcohol and/ or drugs.
Historically, crashes reported to the police have been used as a measure of crash frequencies and types, with the police forwarding crash reports to provincial Motor Vehicle Branches for compilation and statistical analyses. However, a comparison of the frequencies of these reports with data from insurance company crash counts shows an underreporting of less serious crashes in the police-generated data. This could be because of a lack of policing resources, a reluctance on the part of drivers to report crashes to the police (but a desire for financial compensation from insurance companies) or both. Certainly, some proportion of crashes will never be reported to anyone and will just be settled privately, but insurance-based counts seem to gather many more crash instances than do police data counts.
An examination of insurance-based and other data sources suggests that there may be a roughly stable relationships among the number of motor-vehicle related fatalities to the number of injuries to the number of property damage only (PDO) events. For this exercise, for every fatality, there were assumed to be 118 injuries, and 650 PDO events. Using these multipliers, one can move from the very good information on the frequencies of fatalities, to an estimation of the frequencies of less serious crashes.
Similarly, an examination of BAC levels associated with different levels of crash-related injury severity (from no injury to fatality) can produce a rough estimation of the proportion likely impaired by alcohol in less severe crashes for every one percent impaired by alcohol in fatal crashes. An examination of these relationships showed that as crash severity lessened, the likelihood of impairment being a cause lessened. For the purpose of this estimation exercise, the examination of the BAC data suggested that for every one percent of fatal injuries associated with an impaired crash, about half of one percent of injury-only crashes were likely to be associated with alcohol-impairment, and about three-tenths of one percent of PDO events were likely to be associated with alcohol-impairment. To put this another way, if the percent of alcohol-impaired crashes went up by 10%, the percent of alcohol-impaired injury crashes would go up by 5% and the percent of PDO crashes would go up by 3%. Again, using these multipliers, one can move from very good information on the frequencies of impairment-related fatalities to an estimation of the frequency of impairment in less serious crashes.
Finally, an examination of studies of the impairing role of drugs as well as alcohol in crashes suggested that where there is a positive BAC, about 75% of the instances involve alcohol alone, about 25% of the instances where alcohol was involved there were likely also drugs involved, and that there was an additional 10% of persons likely impaired by drugs over and above those impaired by alcohol alone or alcohol and drugs.
Calculation of Frequencies:
In order to estimate the extent of fatal, injury and PDO events, it was assumed that there was 118 injuries and 650 PDO events for every fatal event, and those multipliers were applied to the known number of motor vehicle-related fatal events in Canada in 1999, as reported by the Traffic Injury Research Foundation (Mayhew et al., 2001). From that it was estimated that there were:
Estimated Fatalities, Injuries & PDO Vehicles, Canada, 1999
Fatalities / Injuries @118 / PDO veh. @ 650Total N / 3,315 / 391,170 / 2,154,750
Again, working from Mayhew et al. (2002) the percent of persons killed in motor vehicle-related crashes, on-road or off-road, where alcohol was involved was used as a starting point to estimate the numbers injured and PDO events, using the notion that for every 1% fatal there would be 0.5% injured and 0.3% in PDO events. That resulted in:
Estimated % Alcohol-Involved - Fatalities, Injuries & PDO Vehicles, Canada, 1999
Fatalities / Injuries @0.5% / PDO veh. @ 0.3%% alcohol involved / 34.20% / 17.10% / 10.26%
The next step involved applying the estimations of alcohol-only, alcohol-and-drug, and drug-only percentages to the alcohol-involved percentages.
Estimated Impairment Source - Fatalities, Injuries & PDO Vehicles, Canada, 1999
Fatalities / Injuries / PDO veh.% alcohol involved / 34.20% / 17.10% / 10.26%
% alcohol only @.75 / 25.65% / 12.83% / 7.70%
%alcohol+drug @.25 / 8.55% / 4.28% / 2.57%
%drug only @.10 / 3.42% / 1.71% / 1.03%
% Impaired / 37.62% / 18.81% / 11.29%
Finally, the estimated percent impaired was applied to the estimated number of fatalities, injuries, and PDO vehicles to give an estimated number of victims and PDO vehicles.
Estimated # Impaired-related Fatalities, Injuries & PDO Vehicles, Canada, 1999
Fatalities / Injuries / PDO veh.Total N / 3,315 / 391,170 / 2,154,750
% Impaired / 37.62% / 18.81% / 11.29%
N Impaired / 1,247 / 73,579 / 243,185
The insurance company-generated ratios of 1.2 fatalities per fatal crash, 1.11 injuries per fatal crash, 1.44 injuries per injury crash and 1.52 vehicles per PDO crash were used to move to the crash as the units of analysis[1]. This resulted in an estimation of 1,039 fatal crashes, 50,295 injury crashes and 159,990 PDO crashes associated with impairment by alcohol, alcohol and drugs, or drugs only for 1999.
Using these crash frequency estimations, three costing models were applied (in 1999 dollars) to estimate the order-of-magnitude of impaired-related crashes in Canada in 1999.
Calculation of Costs:
Broadly, there are three kinds of questions that are asked about the result of a traffic crash:
- How much will this cost me in real dollars spent? (Real Dollar Estimate -- RDE)
- How much will this cost me in terms of lost goods, opportunity, or productivity? (Discounted Future Earnings --DFE)
- How much would I pay for this not to have happened? (Willingness to Pay -- WTP)
Each model approaches the question of crash costs differently, especially in the calculation of the value of a fatal crash. The RDE figures are based on estimates from the Insurance Corporation of British Columbia (ICBC), while the DFE and WTP estimates came from an Ontario study by Vodden et al. (1994).
Crash costs by Costing Model in 1999$
fatal / injury-only / PDOReal Dollar Estimate / $280,340 / $25,215 / $1,581
Deferred Future Earnings / $984,412 / $23,779 / $7,265
Willingness to Pay / $7,473,138 / $32,101 / $7,265
Using these models, in 1999 dollars, impaired-related crashes based on the above frequency estimations cost:
Estimated Cost of Impaired Driving Crashes by Costing Models
fatal / injury-only / PDO / SumReal Dollar Estimate / $291,344,243 / $1,268,195,972 / $252,872,971 / $1,812,413,186
Deferred Future Earnings / $1,023,052,566 / $1,195,999,410 / $1,162,332,552 / $3,381,384,528
Willingness to Pay / $7,766,477,399 / $1,614,515,834 / $1,162,332,552 / $10,543,325,785
Conclusion:
A model to give an order-of-magnitude estimation of the extent and cost of alcohol, alcohol-and-drug, and drug-impaired driving crashes in Canada was developed using a range of data sources. Overall, it was estimated that in 1999 there were over 200,000 impaired driving crashes in Canada and that impaired driving cost between $1.8 billion and $10.5 billion, depending upon the costing model used.
However, these figures can only be considered as a rough approximation based on a number of assumptions, any one of which could be in error to a substantial degree. Nevertheless, the issue of the underestimation of less serious crash frequencies and the resulting underestimation of the magnitude of impaired driving crashes and attendant costs is real and worthy of consideration by researchers and policy makers alike.
Finally, it must be emphasized that data on trauma and impairment by alcohol and / or drugs are strongest in the area of traffic crashes, possibly because of the high frequency and financial consequences of these crashes. However, if should be recognized that these kinds of impairment are also likely to contribute substantially to causing trauma in other activities, including boating, snowmobiling, skiing, falls and so on. This paper is restricted to the area of traffic crashes primarily because of access to data of sufficient quality and quantity, not because of an assumption that the frequency and costs of impairment-related trauma from other sources are insignificant.
Introduction
In a climate of limited financial and human resources, proponents of different social issues must compete for support for their issues. Policy makers at all levels of government, the public and private corporate sectors (e.g., the insurance and hospitality industries), and the general public all contribute to some extent to how these limited resources are distributed. Clearly, in this process, it is critical that resource-allocation decisions be made in an informed manner based on data that are as accurate as possible.
There is no question that alcohol, alcohol in combination with drugs, and drugs alone can impair drivers, and that this impairment can lead directly or indirectly to traffic crashes and other sources[2] of trauma and mortality. There is also no question that this situation takes a serious and substantial toll in both human and financial terms. To illustrate, in 1998 in Canada, there were 555 homicides reported to the police (Canadian Center for Justice Statistics, 1999), while in the same year 2,909 persons were killed in traffic crashes and 986 of those involved a drinking driver (Mayhew et al., 2001).
The precise measurement of the contribution of this impairment to the frequency and severity of crashes, as well as their cost to the social system, are more difficult questions to address. Data sources, data definitions, units of analysis, analytic techniques, and sociopolitical perspectives and interpretations differ across agencies, jurisdictions, and analysts.
This is not to imply that any particular statistical analysis is in error, or that advocates of one policy or another intentionally bias their reports, but, rather, that different approaches naturally produce different results. Having said that, there is a need to explore whether a model for estimating a global order-of-magnitude of the extent and cost of impaired driving can be developed that takes into account the strengths and weaknesses of the various data sources and that could be generally used in this area.
Consequently, it is the purpose of this report to begin a dialogue on what sort of model might be appropriate for such a global estimation of the extent and cost of impaired driving in Canada.
An Overview of Some Common Crash Data Sources and their Limitations
Crash Severity & Frequency Data Sources
It is impossible to know with absolute certainty the number of traffic crashes that occur in a jurisdiction. The reason for this is that the data sources vary in how they define crashes and how they compile their information. For example, consider three common sources of crash data: data reported to Motor Vehicle Branch by the police, motor vehicle insurance record data, and coroner data. Table 1 represents the number of persons reported killed in British Columbia in motor vehicle crashes in 1998 drawn from these three sources, while Table 2 compares police reports to insurance company counts on the number of fatal crashes, injury crashes, and property damage only (PDO) crashes in BC and in Manitoba in 1998. There are a number of reasons why these numbers differ[3] but the point is that the same question can result in very different responses, depending upon the data source queried.
Table 1: Motor Vehicle-Related Fatalities, BC, 1998
# and Data SourcePolice / Insurance (ICBC) / Coroner
Crash Fatalities / 418 / 473 / 455
Table 2: # Motor Vehicle Crash Types, BC & Manitoba, 1998
Crash # and Data SourceBritish Columbia / Police / Insurance (ICBC)
Fatal / 365 / 418
Injury / 17,984 / 46,554
Property Damage Only (PDO) / 17,418 / 203,245
Total / 35,767 / 250,217
Manitoba / Police / Insurance (MPI)
Fatal / 109 / 138
Injury / 6,879 / 27,124
Property Damage Only (PDO) / 20,136 / 60,239
Total / 27,124 / 69,003
Police Data:
Most jurisdictions collect data on traffic crash frequencies from police reports, but these reports probably underestimate the actual number of crashes, especially those that are less severe.
For a crash to be included in police crash counts, at least three things must happen: the crash must be reported to the police; the police must attend the crash; and a report must be written and submitted to the compiling authorities (usually Motor Vehicle Branch in a Ministry). Admittedly, some jurisdictions have allowed motorists to submit a report to the police, eliminating the second step of police attendance, but the reliability and validity of self-reports can be questionable.
Considering the first step of reporting a crash to the police, there is the disincentive for some drivers to report a crash for fear of precipitating Motor Vehicle Act or even Criminal Code charges, as well as avoiding civil liability. In a meta-analysis of crash reporting, Elvik & Mysen (1999) found that about 95% of the fatal crashes, 70% of the serious injury crashes, 25% of the slight-injury crashes, and 10% of the very slight injury crashes were reported to the police.
Once a crash has been reported to the police, the pressures of other calls and shortage of personnel can make it impossible for the police to attend. Further, the police are probably more likely to attend more serious crashes, crashes that require traffic control (e.g., intersection), crashes that occur during times of day and days of the week that have lighter call loads, and / or those where charges might be more likely to be laid, resulting in one or more biases in the data.
Finally, even when a crash has been attended, there is no guarantee that a crash report will be completed and filed appropriately.
One way of getting some idea of the magnitude of difference between police counts and actual crash numbers is to compare police counts to crashes as recorded by motor vehicle insurance agencies (although the problem of unreported crashes remains for both sources). In British Columbia all vehicles must be insured through the Insurance Corporation of British Columbia (ICBC), while in Manitoba they must be insured through Manitoba Public Insurance (MPI). Using data from these Crown Corporations and the respective police data, in BC in 1999 police attended and reported on 375 of the 405 fatal crashes known to ICBC (92.6%), 19,947 of the 45,077 injury crashes known to ICBC (44.3%), and 20,987 of the 206,567 property-damage-only (PDO) crashes known to ICBC (10.2%), while in Manitoba the 1999 proportions were 75% of the 132 MIP fatal crash count, 76% of the 9,175 injury crash count and MPI and 35% of the 62,528 MPI PDO crash count[4].