Annex 4 analytical methods

model description

general description

acronym: VeSTEM

name: Vehicle Safety Technology Effectiveness Model

main purpose: The VeSTEM model allows predicting casualty savings and costs arising from the simultaneous implementation of multiple vehicle safety systems and their interactions in preventing and mitigating of collisions. Model outputs allow evaluating the cost-effectiveness and casualty prevention potential of different sets of systems for the assessment of policy impacts.

homepage: https://publications.europa.eu/en/publication-detail/-/publication/ed4aff17-49c5-11e8-be1d-01aa75ed71a1/language-en

Developer and its nature

ownership: Third-party ownership (commercial companies, Member States, other organisations, …)

ownership additional info: TRL Limited, UK

is the model code open-source?: NO

Model structure and approach with any key assumptions, limitations and simplifications

details on model structure and approach

VeSTEM is a model to estimate the benefits (monetary values of casualties prevented by safety measures) and costs (cost to vehicle manufacturers of fitment of safety measures to new vehicles) associated with policy measures assessed in the context of the revision of the General Safety Regulation and Pedestrian Safety Regulation. The model is implemented in the programming language Python (https://www.python.org/) with inputs and outputs produced in Microsoft Excel spreadsheets.

The model considers as benefits the monetary values of casualties prevented by safety measures, and as costs the cost to vehicle manufacturers of fitment of safety measures to new vehicles. Results are benefit-to-cost ratios (BCRs), based on present monetary values and casualties prevented, compared to the baseline scenario over the entire evaluation period.

A vehicle fleet calculation module determines how the vehicle safety measures disperse into the fleet. The module determines the effect of mandating a measure for all new types, and two years later for all new registered vehicles, on the overall proportion of the fleet equipped. Benefits conferred by a safety measure, that is, casualties prevented, will only be realised by equipped vehicles. However, the legacy fleet will also be affected by active safety measures; for example, if a rear-end shunt is avoided by advanced emergency braking for driving and still-standing vehicles ahead, the vehicle in front, will benefit from the measure even if it is a legacy vehicle. This is taken into account in the benefit calculations.

To simulate the casualties prevented by each measure, an accident data analysis was performed based on Great Britain national road accident data (Stats19) to determine the casualty target population for each proposed measure (input data), i.e. the number of fatal, serious and slight injuries that could potentially be affected by a safety measure based on relevant characteristics of the collision (e.g., collision geometry or contributory factors). The target populations were scaled to EU-27 and UK level using weighting factors, based on severity and vehicle categories involved, derived from analysis of the pan-European CARE database. The target populations found are multiplied with effectiveness values for each safety measure (input data), i.e. a percentage value indicating what proportion of the relevant accidents will be avoided or mitigated by the measure. Mitigated casualties (fatal turned to serious casualty, or serious to slight casualty) are added to the target population of the next lower injury severity level for other measures. The casualties prevented are multiplied with monetary values for casualty prevention to calculate the monetary benefit.

The model also addresses the interaction of different safety measures on overlapping casualty groups. To give an example, there are collisions where a driver was exceeding the speed limit, left the lane and suffered a frontal impact. These collisions will be in the target populations for multiple measures, but they can only be prevented once by either one of these systems. This is addressed in the model by removing casualties prevented by one measure from the subsequent target population of the other measures. The impact of highly effective existing safety measures, which have been mandatory for a few years, but are still dispersing into the vehicle fleet is also modelled to reduce the remaining target populations for the proposed measures.

The cost of a policy option is calculated by multiplying per-vehicle cost estimates (input data) for each measure with the number of new vehicles of each vehicle category across EU-27 and UK that are equipped with the measure in the given year of the analysis according to the output of the fleet calculation model. In the economic calculation module, the monetary values of costs and benefits are subjected to inflation and discounting to determine their present value. The present values of benefits and costs exceeding the baseline, calculated for individual years and summed over the study period, are compared in order to arrive at cost-effectiveness estimates.

model inputs

Benefits considered:

monetary values of casualties prevented by safety measures

Costs considered:

cost to vehicle manufacturers of fitment of safety measures to new vehicles

model outputs

Results:

Number of fatal, serious and slight casualties prevented
Benefit-to-cost ratios (BCRs), based on present monetary values and casualties prevented, compared to the baseline scenario over the entire evaluation period.

Intended field of application

policy role: VeSTEM can be used to model the future dispersion of vehicle safety systems into the EU fleet and quantify the EU-wide number of casualties prevented by voluntary or mandatory implementation of the systems. The model also allows to monetise the casualty savings (benefits) and calculate fitment costs associated with the systems (cost) and perform cost-effectiveness calculations for policy options including different sets of safety measures.

policy areas

Transport

Model transparency and quality assurance

Are uncertainties accounted for in your simulations?: YES - Input parameters having a relatively high associated uncertainty were identified and given upper and lower bounds of variation. These were used to calculate variation in the BCR (from absolute lower BCR to absolute upper BCR).

Has the model undergone sensitivity analysis?: YES - Two sensitivity analysis techniques were used: An interval and a scenario analysis were carried out to quantify the range of uncertainty around the best estimate BCR values.

Has the model been published in peer review articles?: NO

Has the model formally undergone scientific review by a panel of international experts?: NO

Has model validation been done? Have model predictions been confronted with observed data (ex-post)?: YES - The model has been verified during the course of implementation using dummy inputs and verifying the individual calculation steps data, but the model has not been validated ex-post.

To what extent do input data come from publicly available sources?: Based on both publicly available and restricted-access sources

Is the full model database as such available to external users?: NO - Collision databases used to create input data for the model (STATS19 and CARE) can be accessed at request to the database owner. Other input data was collected from literature and stakeholders and is provided in the project report (see References).

Have model results been presented in publicly available reports?: YES

Have output datasets been made publicly available?: NO - Model results were made publicly available in the project report.

Is there any user friendly interface presenting model results that is accessible to the public?: NO

Has the model been documented in a publicly available dedicated report or a manual?: YES - General model structure, input data and results are documented in a project report.

Intellectual property rights

Licence type: Non-Free Software licence

application to the impact assessment

Please note that in the annex 4 of the impact assessment report, the general description of the model (available in MIDAS) has to be complemented with the specific information on how the model has been applied in the impact assessment.

See Better Regulation Toolbox, tool #11 Format of the impact assessment report).