Crash Costs

Utah uses the KABCO injury classification scale when reporting crash injury level. This scale has been modified to a 5-1 scale for data storage purposes. Eventually the crash severity name is applied for reporting purposes as shown below.

• K = 5 = Fatal

• A = 4 = Suspected Serious Injury

• B = 3 = Suspected Minor Injury

• C = 2 = Possible Injury

• O = 1 = No Injury/Property Damage Only

Crash costs are based on the Maximum Abbreviated Injury Scale (MAIS) that is used by doctors in diagnosing injury severity. Conversion factors provided by the United States Department of Transportation (USDOT) are used to convert the MAIS to the KABCO scale. This discussion refers to the KABCO scale in order to conform with applicable national publications. 

Key Principles

It is important for engineers and analysts to understand the following key principles related to crash costs:

1. The cost associated with a fatal crash is not an attempt to declare the value of human life. Instead, fatal crash costs are a function of the value of statistical life (VSL). The USDOT defines the VSL as, “the cost that individuals would be willing to bear for improvements to safety...that in the aggregate, reduce the expected number of fatalities by one.” This value is associated with the reduction of fatality risk, not the value of a human life.

2. Crash costs are a function of two values:

   • Quality-Adjusted Life Years (QALY): the, “intangible consequences-such as the physical pain and emotional suffering of people injured in crashes and their families.” The USDOT provides factors of the VSL to determine the QALY for non-fatal crashes. 

   • Economic crash costs: estimates of direct and indirect economic costs to individuals and society relating to a crash. 

3. Crash costs are AVERAGES that account for a variety of situations. 

4. Crash costs are based on an understanding of the statewide average severity distribution (i.e. the proportion of crashes by severity level).

Frequently Asked Questions

1. Do crash costs account for multiple deaths/injuries in a single crash?

   • Yes, values are adjusted based on the typical ratio of crashes to injury severity.

2. Do crash costs account for societal costs like traffic delays, emergency vehicle costs, etc?

   • Yes, these are part of the economic crash costs.

3. Do crash costs account for individual costs like vehicle repair, insurance claims, lost wages etc?

   • Yes, these are part of the economic crash costs.

4. Are crash costs the same for all crash types?

   • Yes, crash costs are calculated based on the AVERAGE statewide severity distribution including all crash types. Refer to the Final Considerations section below for additional discussion.

5. Do crash costs include unreported crashes?

   • Yes, the National Highway Traffic Safety Administration found that approximately 60% of all PDO crashes and approximately 40% of all non-fatal injury crashes were unreported. These additional crashes are accounted for in the crash cost estimates.

Which Crash Costs Should I Use?

Because different analysis methods generate crash predictions based on a variety of severity groupings (for example, KAB, or KABC, or any other combination), it is important to use the weighted crash costs that best suit the analysis method. Table two presents a few common scenarios:

Analysis Method:

When using a predictive model from the Highway Safety Manual (HSM) where there is no severity distribution

Weighted Crash Costs:

Chapter 10 of the HSM provides Safety Performance Functions (SPF) for fatal and injury crashes (FI, or “KABC” on the KABCO scale), property damage only crashes (PDO, or “O” on the KABCO scale), and total crashes (“KABCO” on the KABCO scale). When using SPFs to predict the FI and PDO crash frequencies, the analyst should use the weighted KABC crash cost and the O crash cost to calculate the safety benefit (refer to current crash costs, Table 1).

Note: If you have a Utah specific severity distribution for, say, rural two-way two-lane highways, then the crash costs would not be weighted. Rather, you would apply the individual crash costs for K, A, B, C, and O for each predicted value.

Analysis Method:

When using a predictive model from the HSM where there is a severity distribution

Weighted Crash Costs:

Chapter 12 of the HSM provides SPFs for various crashes along with the severity distribution for the frequencies calculated. Because the severity distribution is available, no weighting of the crash costs needs to be applied when calculating the safety benefit, apply the individual unweighted crash costs for K, A, B, C, and O for each predicted value.

Analysis Method:

When preparing an HSIP application using the severity distribution of the observed crash history

Weighted Crash Costs:

Historically, the observed crash frequency is assumed to represent the severity distribution of the location that is evaluated. For this analysis the Severe crash cost is used to help “dilute” the value of the fatal crashes (refer to current crash costs, Table 1). All other crash costs are applied to the specific severity of the observed crashes. Furthermore, only the targeted crashes are considered in the analysis, and the CMFs for the specific measure is applied. The analyst should recognize the weakness of these assumptions:

1. The analysis method is particularly susceptible to regression to the mean since HSIP applications are typically identified due the occurrence of a severe crash. 

2. Using the weighted crash costs for both fatal and suspected serious injury crashes dampens the effect of fatal crashes, it also over emphasizes the value of suspected serious injury crashes when calculating the safety benefit. 

3. The error is exaggerated for locations with very low crash counts because the observed crashes assume the crash severity distribution will stay constant over the analysis period.  For example,having only one fatal crash at a rural intersection in three years, does not mean there will be another fatal crash in the next 3 years. 

Note that using severity distributions for various segments and intersections in Utah and unweighted crash costs is preferred.

Final considerations

Analysts should be familiar with the assumptions that are applied to each analysis method and how those assumptions affect crash costs. Some final thoughts engineers and analysts should remember:

1. Crash costs are based on the severity distribution of all crash types. Some crash types are inherently less severe than the state average distribution (e.g., wildlife crashes, winter weather related crashes, etc) while others are more severe than the state average (e.g. pedestrian crashes, roadway departure crashes, angle crashes). In these cases the analyst should keep in mind that the average crash costs may over-  or under-represent the safety benefit of addressing these crashes. 

2. Typically crash costs are established by severity and not by manners of collision. This is done for convenience and avoids the potential confusion of why one manner of collision might have higher crash costs for the same severity of crash than another manner of collision. A multi-vehicle crash will almost always be more expensive, but we average crash costs and treat them as the same for simplicity.

3. Predictive models assume that engineering design and judgment have been applied equally to all situations of the roadway. This assumption also applies to severity distributions. However, the analyst might identify a location where some element of design or operation is leading to a disproportionate crash severity distribution or manners of collision frequency (e.g. a traffic signal that would meet the warrant for protected turn phasing but has not been evaluated and upgraded as such). Glossing over these locations with predicted averages might minimize the opportunity to make significant improvements to safety. While predictive models are still the preferred method, flagging these observations and sharing them with decision makers is important.

4. The methodology outlined in HSM chapters 11-12 applies severity distributions and manner of collision proportions as multipliers to the total crashes. This assumes that the severity distribution of one manner of collision is equal to the severity distribution of another manner of collision (i.e. the severity distribution of angle crashes is the same for side-swipe-same direction crashes). In reality this is not the case. The analyst should always keep in mind the severity distributions and their relationship to the crash costs and how that may impact the final results of the analysis.