A6.3 - Applying the analysis methods
- A6.1- Accident costs
- A6.2 - Choosing to undertake an accident analysis
- A6.3 - Applying the analysis methods
- A6.4 - Accident trends
- A6.5 - Typical injury accident rates and prediction models
- A6.6 - Typical accident reduction factors
- A6.7 - Adjusting accident costs to reflect mean speeds
- A6.8 - Worked example of accident procedures
- A6.9 - Tables
- A6.10 - References
A6.3 - Applying the analysis methods
Introduction
This section describes the general process for how to determine future annual accident numbers and costs for the do minimum and project options using the three analysis methods:
- Method A: Accident-by-accident analysis
- Method B: Accident rate analysis
- Method C: Weighted accident procedure.
Worked examples of the Methods B and C are provided in appendix A6.8.
Intersection accidents
Accidents occurring within the area of priority controlled intersections, roundabouts and traffic signals on the primary road network, and up to 50 m from the intersection.
Mid-block accidents
Accidents occurring on a road section excluding accidents at major intersections. Accidents at minor intersection are sometimes included.
Categorisation by speed limit
Accidents are categorised according to the speed limit areas in which they occur:
- 50 km/h speed limit areas (including 30 km/h and 60 km/h areas)
- 70 km/h speed limit areas (including limited speed zones)
- 100 km/h speed limit areas (including 80 km/h and above areas).
Types of accident rate
An accident rate is the average number of injury accidents per year, measured over a period of time (normally five calendar years).
Site-specific accident rate (AS)
is the accident rate for a specific site based on reported injury accidents on the record of TCRs prepared by the Police and compiled by Land Transport NZ. These are available from the Crash Analysis System (CAS).
Typical accident rate (AT)
is the accident rate for a typical or generic site, eg, a bridge, with characteristics similar to the site being evaluated. Typical accident rates are determined using either an accident prediction model or exposure-based accident prediction equation, depending on the type of site, or part of a site, being evaluated.
Weighted accident rate (AW)
The accident rate produced when using the weighted accident procedure.
Method A: accident-by-accident analysis
Accident-by-accident analysis is based on the accident history of the site and is dependent on the number of reported accidents, as set out in appendices A6.1 and A6.2. The analysis uses the individual accident severity categories (fatal, serious, minor, non-injury) and these can be further disaggregated by movement category and/or type of vehicle involved.
In the first stage of the analysis, using the worksheets in chapter 5, the do minimum total estimated number of accidents per annum is calculated. Costs are assigned using the accident costs from tables A6.21(a) to (d) for 50 km/h speed limit areas and from tables A6.21(e) to (h) for 100 km/h speed limit areas.
The number of accidents predicted for a project option is determined from an expected reduction in the do minimum accident numbers, based on the guidance provided in Appendix A6.6. The forecast percentage accident reductions for the project option can be applied either globally or varied for each accident type and severity (eg, for fatal, serious, minor and non-injury accidents). Costs are taken from tables A6.21(a) to (h) as appropriate to the site. Where the mean speed of traffic for the do minimum and/or options differs from that provided in table A6.21, an adjustment should be made to the costs using the formula found in appendix A6.7.
Severity
In Method A, accidents are categorised by the most severe injury sustained. The four severity categories are:
| Fatal | When death ensues within 30 days of the accident. |
| Serious | Injuries requiring medical attention or admission to hospital, including fractures, concussion and severe cuts. |
| Minor | Injuries other than serious, which require first aid or cause discomfort or pain, including bruising and sprains. |
| Non-injury | When no injuries occur, sometimes referred to as 'property damage only' (PDO) accidents. |
The accident reports from police officers recorded in CAS are to be used to classify accident severity in preference to hospital records.
Changes in accident severity
Project options, such as crash barriers, can in some cases reduce the accident severity at a site. Use Method A, rather than Method B or C, when the majority of accident benefits are obtained from a reduction in accident severity.
Vehicle involvement
In assigning costs to accidents using Method A, accidents are classified by 'vehicle involvement' according to the highest ranked 'vehicle' involved in an accident. The ranking from highest vehicle to lowest vehicle is:
- pedestrian
- push cycle (bicycle)
- motorcycle including moped
- bus
- truck
- cars, light commercial vehicles and any other.
For example, an accident involving a truck and a push cycle is categorised as a 'push cycle accident'.
Adjustment for under-reporting
Only a proportion of non-fatal accidents that occur are recorded on TCR and in CAS. This is referred to as under-reporting. It is generally assumed that all fatal accidents are reported.
To counteract the effect of underreporting when using Method A, factors are applied to reported accident numbers (TCR numbers) to estimate the total number of accidents that actually occur. Table A6.20(a) provides factors for converting from reported injury accidents to total injury accidents, while table A6.20(b) provides factors for converting from reported non-injury accidents to total non-injury accidents.
If local contact accident information has been used, then under-reporting factors must not be included in the calculations of injury or non-injury accident costs.
Change in traffic volume
If there is a change in traffic volume for the project option compared with the do minimum, then the accident numbers must be scaled in proportion to this change.
Method B: Accident rate analysis
Accident rate analysis involves determining a typical accident rate per annum as the basis for calculating the accident cost savings for a project. Typical accident rates have been calculated using either an accident prediction model or an exposure-based accident prediction equation from appendix A6.5, which have been derived using information from similar types of site elsewhere.
In some cases, the models used for the do minimum and the project option already account for the proposed improvement/treatment of the site (eg, an intersection treatment to change from priority or a roundabout to signalised; the construction of a two-lane rural bridge to replace a single lane bridge). In others, it may be necessary to apply an accident reduction factor from appendix A6.6 to the project option model or equation to take account of the site treatment/improvement (eg, various mid-block pedestrian treatments; construction of a cycle lane).
In accident rate analysis, it is not possible to differentiate accidents other than by speed limit category, therefore the accident costs are taken from table A6.22, and are for 'all vehicles and all movements combined'. Where the mean speed of traffic for the do minimum and/or options differs from that provided in table A6.22, an adjustment should be made to the costs using the formula found in appendix A6.7.
Only reported injury accidents are considered when using accident rate analysis because of the inconsistency in non-injury reporting rates from district to district.
Method C: Weighted accident procedure
The weighted accident procedure uses both historical accident data relating to a particular site, and the typical accident rate for the site, as calculated from the appropriate accident prediction model or exposure-based accident prediction equation (from appendix A6.5).
The historical data is converted into a site-specific accident rate by dividing the reported accidents by the number of years of data. The site-specific accident rate is then combined with the typical accident rate, resulting in a weighted accident rate for the do minimum and project option(s).
Accident cost savings for the do minimum and option(s) are calculated using the costs provided in and table A6.22. Where the mean speed of traffic for the do minimum and options differs from that provided in table A6.22, an adjustment should be made to the costs using the formula found in appendix A6.7.
The weighted accident procedure also allows analysis of sites with no accident history, provided that the site has been in existence for more than 3 years with no major changes.
Use of site specific accident rates
For existing links, use site-specific accident rates calculated from the accidents that have occurred on the links.
Where there is low accident occurrence due to short link lengths or low traffic volumes, site-specific accident rates can be unrealistic. In this case, accident prediction models, exposure-based accident prediction equations or site-specific accident rates from adjoining links should be used to determine future accident numbers. Intersections and other sites can be similarly analysed if necessary.
Weighted accident rate for the do minimum
The do minimum weighted accident rate is calculated using the following equation:
AW,dm = w × AT + (1 − w) × AS
where:
AW,dm is the do minimum weighted accident rate
AT is the typical accident rates calculated from the appropriate accident prediction model or exposure-based accident prediction equation (from appendix A6.5) for the do minimum
AS is the site-specific accident rate (from historical accident data)
w is the weighting factor.
Weighting factor (w)
When w = 1, the method simplifies to an accident prediction model or equation (Method B).
When w = 0, the method simplifies to an accident-by-accident analysis (Method A).
w is calculated using the following equation if k is specified:
If k is not specified then the equation for w is:
Where: k is a dispersion parameter (defined below), and
aX and aM are reliability factors (defined below).
Dispersion parameter (k)
k is a dispersion parameter of the negative binomial distribution, which is the probability distribution assumed for the accident data. k values for different sites are in appendix A6.5.
Generally the higher the value of k the higher the accuracy of an accident prediction model (and vice versa). The accuracy is, however, also relative to the typical accident rate at a site, ie, a low k value may be acceptable at a site with a low typical accident rate but unacceptable at a site with a high typical accident rate.
For a mid-block, the typical accident rate (AT) must be divided by the length of the mid-block because the mid-block k values provided in appendix A6.5 are on a per kilometre basis.
Reliability factors (aX, aM)
An assessment of the reliability of both the site-specific accident rate and the typical accident rate is required for Method C. The reliability factor for the site-specific accident rate is aX and the reliability factor for the typical accident rate is aM.
The main factor influencing the reliability of the site-specific accident rate is whether accidents are correctly coded at the site. Accidents may be missing from the site or may be incorrectly coded within the site. For example, an accident may be incorrectly coded within a series of back-to-back curves, where it is not always easy to accurately locate the exact curve the accident occurred on.
When the historical accident data is reliable, aX should equal 1.0 (this is the default setting). When it is unreliable, aX should be between 1.0 and 2.0, with 2.0 being very unreliable data.
The reliability of the typical accident rate information presented in appendix A6.5 is an issue when an accident prediction model or exposure-based accident prediction equation is used for:
- A different type of site, or part of a site, than the model or equation was derived for. For example, a 4-arm roundabout model might be used for a 3-arm roundabout (the prediction would then be approximately 75 percent of that given by the model)
- A 'non-standard' intersection, mid-block or other site or part of a site. An example of a 'non-standard' intersection would be one with many traffic signal phases (say 5 or 6) or greater than four approach lanes.
In both situations aM should be increased above 1.0 (the default value). A value of 2.0 would represent poor reliability.
Weighted accident rate for project option
Method C can only be used for the project option when it does not bring about a fundamental change in a site. In this case, the site-specific historic accident data is still relevant for the project option. The project option weighted accident rate is calculated by increasing or decreasing the typical accident rate of the project option by the same proportion used to adjust the do minimum typical accident rate to the do minimum weighted accident rate.
AW,opt = AT,opt × AW,dm / AT,dm
where: AW,opt is the weighted accident rate for the option
AW,dm is the weighted accident rate for the do minimum
AT,opt is the typical accident rate calculated from accident prediction models or exposure-based accident prediction equations for the option. Note that it may be necessary to apply a reduction factor from Appendix A6.6 if the prediction model or equation does not already take the treatment / improvement into account.
AT,dm is the typical accident rate calculated from accident prediction models or exposure-based accident prediction equations for the do minimum.
