The objective of this Report is the analysis of the state-of-the-art in risk and exposure data availability, collection methodologies and use in the European Union. More specifically, the analysis aims to explore the concepts of exposure and risk, as well as the theoretical properties of the various exposure measures in use in road safety. Moreover, it aims to present an overall picture of the existing methods for collecting exposure data for national risk estimates. Finally, the potential of international risk comparisons is investigated, mainly through the International Data Files with exposure data.
In order to meet these objectives, the following methodology was adopted: firstly, an exhaustive bibliography review was carried out and a bibliography database on risk and exposure data was developed. Additionally, a set of National Reports was created by the institutes involved in the analysis, providing representative examples of exposure data availability, collection methods and use from seven representative European countries: Denmark, France, Greece, Hungary, Norway, the Netherlands and Portugal. Furthermore, a separate survey was devoted to the investigation of the International Data Files, as far as exposure data availability and quality is concerned. The survey was carried out by means of personal interviews with the maintainers of the related databases of the following organizations: EUROSTAT, ECMT, UNECE, IRTAD and IRF.
From the results of the analysis, it was deduced that, comparing risk rates, especially at international level, may be a very complex task. Both accident counts and exposure measures present some theoretical and practical limitations and are subject to estimation errors, which may compromise their usability. Especially as far as exposure is concerned, in theory, continuous exposure measurements of different road user categories in different modes and different road environments would be required and could provide detailed exposure estimates to the degree of disaggregation of the respective accidents data. In practice, such measurements are not possible, therefore, road safety analyses need to compromise to some approximations of the actual exposure, which may be more or less accurate and representative. Different exposure measures may be used, according to data availability and quality, as well as the context of the analysis. It should be noted that no general rule can be adopted on the preferred measures of exposure.
However, it can be deduced that the most appropriate measures of exposure appear to be vehicle- and passenger-kilometres of travel, because they are closer to the theoretical concept of exposure and can be available, in theory, to a satisfactory level of detail. However, other exposure measures are often used, namely the vehicle fleet and the drivers' population, the road network length, the fuel consumption, as well as the entire population, mainly because they involve less complex collection methods.
The theoretical features of the various exposure measures were analyzed in detail in the framework of the present research. In practice, however, the availability, quality and disaggregation level of exposure measures may be compromised by limitations and particularities of the respective collection methods. The main sources of exposure data include travel surveys, traffic counts systems, vehicle fleet register, driving licenses registers, roads registers and population registers.
Travel surveys are carried out in most European countries, in order to collect information on traffic and mobility patterns. From the data collected (namely distance traveled, time spent in traffic and number of trips), vehicle- (actually driver-) and passenger-kilometres estimates can be obtained. The main advantage of national travel surveys (compared to other collection methods) is that these surveys have persons as a unit, making it possible to compare groups of persons, and are usually designed to achieve a high level of data disaggregation by person, vehicle and road network characteristics. However, travel surveys are sample-based self-reporting information collection methods, consequently a number of possible biases (sampling, non response or measurement errors) may occur and should be treated accordingly where possible.
On the other hand, in most countries traffic counts systems are in place, providing data on traffic volumes, which are used to obtain vehicle kilometres estimates. An important advantage of using this method is that the seasonal variations of exposure can be captured, as the measurements are usually continuous over time. However, this method is does not allow distributing exposure by to person characteristics. Additionally, this method is also sample-based, in the sense that measurement points are placed on specific sections of the main road network, which may or may not be representative of the entire road network, and usually local or urban roads are not included. Problems may also be encountered in the automatic classification of vehicles.
The two methods discussed above present different advantages and limitations, however they are the only methods that can produce detailed vehicle- and person-kilometres estimates. However, because of the difficulties in the implementation and operation of such systems, in most countries the vehicle fleet and driving licenses registers are also used to calculate alternative exposure measures. The problem when using such registers to estimate risk is that these are certainly very crude estimates of exposure, giving quite uncertain risk estimates. It should be noted that, data from such databases are known to lead to some (but often uncalculated) overestimations, due to insufficient updating of the registers.
Accordingly, roads registers are often used to apply the length of roads as an exposure measure. However, in most countries the available information concerns the main road (motorways, national and rural roads etc.), whereas information on roadway geometry is less available, and regional/local road length estimates are less available.
From the analysis of examples of implementation of the above methods in the selected European countries, the following conclusions can be drawn:
- The features and specifications of each method may vary significantly among countries
- Accordingly, the availability, disaggregation and comparability of exposure measures (in terms of definitions, variables and values) is quite diverse.
- The disaggregation level theoretically possible for an exposure measure is seldom achieved in practice
- Data from different sources (collection methods) are often used to produce a national exposure estimate, i.e. different data sources may function complementarily for the calculation of a single exposure measure
- In general, it is not always clear how the exposure estimates are obtained from the "raw" data collected by means of the various methods.
- According to the above, it can be deduced that the national exposure and risk estimates may not always be comparable at EU level.
However, in most countries some national exposure estimates are available, which are collected, exploited and published through the International Data Files (IDFs) in the field of transport and road safety. The main IDFs involved in road accident and exposure data EU are the following: Eurostat, ECMT, UNECE, IRTAD and IRF. These data files are useful and accessible aggregate data sources, as a result of several decades of important data collection efforts. However, they have different objectives; they collect different data in different forms and structure, in some cases by different national sources, and are maintained by organizations with different scopes and policies. Consequently, the availability of exposure data among the data files varies significantly, in terms of both countries and years available, and variables and values available.
In the framework of the present analysis, a detailed comparison of exposure data published by the IDFs was carried out, in terms of availability and quality, and several interesting results and conclusions were obtained:
- The exposure data available in the IDFs are in a much more aggregate form than the exposure data collected at national level
- Accordingly, the more disaggregate national exposure data are not exploited within the context of IDFs.
- Significant differences are observed among the IDFs in the published figures for each exposure measure; these differences are more important for the more "sophisticated" exposure measures (i.e. vehicle and passenger kilometres).
- These differences are partly due to the different national sources and definitions used
- However, another reason may concern insufficient data quality control within the IDFs.
Summarizing, the availability and quality of risk exposure estimates in the EU Member States varies significantly, and is related both to the exposure measures used and the characteristics of the respective collection methods. In particular, significant efforts are made at national level to improve data availability, disaggregation and reliability. However the lack of a common European framework for the collection and exploitation of RED limits significantly the comparability of the detailed national data. On the other hand, the International Data Files containing road safety related data, including RED, provide useful aggregate information in a systematic way and are currently the only sources allowing international comparisons, however more effort is required to further improve the availability and quality of these data.
It can be deduced that a series of problems, namely poor data availability, insufficient reliability, inappropriate disaggregation and limited accessibility are the main limitations to the full exploitation of risk and exposure data at European level. It is also obvious, from the analysis presented in this Report, that the most useful RED are the least available. Further work and research should focus on improving data compatibility and availability, namely through a common framework including common data requirements, definitions and collection methods.
In particular, from the results of the state-of-the-art survey on risk and exposure data, which was carried out in the framework of the present research, the following recommendations are suggested, towards a common risk exposure data framework:
- Priority should be given to the collection of vehicle- and person-kilometres of travel, these measures being the most appropriate exposure measures in the context of road and traffic safety analysis.
- The common framework should focus on the collection of disaggregate time series of exposure by road user, mode and network characteristics, and should be organized to provide data in a consistent and systematic way.
- Consequently, both travel survey and traffic counts methods should be exploited, allowing for flexibility, high level of disaggregation and continuity over time in the exposure estimates.
- Additional data sources could be exploited to benchmark or validate the exposure estimates and improving data reliability and accuracy
- The specific calculation process of exposure measures should be defined and standardized.
Certainly, the establishment and application of such a common framework would be a very complex and time-consuming task, which would also involve a significant effort and cost, both at national and EU level. However, given the importance of an improved risk and exposure data availability and quality, to support and monitor an efficient road safety policy at EU-level, it is necessary to promote its development.
However, in order to deal with the current needs, the gathering and harmonization of the existing information is equally important for the improvement of the exploitation potential of the existing exposure data. The harmonization of the definitions of exposure measures, variables and values between countries (at the most disaggregate level), as well as within the International Data Files (aggregate level), in accordance to the existing accident data, as well as the current and future exposure data needs, would be an important first step to improve comparability of the existing disaggregate data.