LGRAS (00) 6 - DRIVER AND CASUALTY POSTCODES, AND THEIR USE IN A RESEARCH PROJECT ON TOURIST ROAD ACCIDENTS IN RURAL SCOTLAND
1. This paper has two parts. First, it describes the current position on the supply, in the "STATS 19" returns, of the postcodes of drivers' and casualties' home addresses, and invites views on the likelihood of biases and the possibilities for improvement. The second part sets out briefly the details of a research project to be carried out on tourist road accidents in rural Scotland. Those involved in recording injury accident data are asked to be particularly alert to the need for as accurate information as possible on the postcode and foreign driver status of the drivers involved in road accidents in all of Scotland over the coming summer months.
The supply of Driver and Casualty Postcodes in the "STATS 19" returns
2. The 1997 Quinquennial Review of the "STATS 19" system recommended that, with effect from January 1999, the postcode of the person's home address should be included as part of the statistical records for each of the drivers and the casualties who were involved in an injury road accident. The driver and casualty postcodes (and the use of the special code for non-UK residents) are potentially of great value in the analysis of the road accident statistics data. For example:
- they will be used in the forthcoming research project on tourist road accidents in rural Scotland, which is described briefly below;
- they will allow one to see where people who are involved in road accidents in certain areas and/or on particular routes have come from, and so assist the implementation and/or targetting of road safety measures;
- they will enable analyses to use "socio-economic" background information about the kinds of areas in which people involved in road accidents live;
- they would increase the range of statistics that is available for areas within Scotland, for example one could produce tables showing the approximate numbers of casualties by Council area of residence, which would permit the calculation of more meaningful casualty rates (traditionally, "STATS 19" could provide only the numbers of casualties in accidents which occurred in each Council area).
3. The "STATS 20" instructions for the completion of the returns mention that part postcodes (eg "EH10 ") may be entered if the full postcode is not known. "STATS 20" also specifies the codes that are to be used in cases where none of the postcode is known, where the person is not a UK resident, and where a vehicle was parked and unattended.
4. The introduction of the supply of postcode information was accompanied by "teething troubles" in a few cases: the percentages of drivers and casualties for which "informative" postcodes were provided was very low for a few Police Forces for a few months. (For the purposes of this paper, an "informative" postcode is defined as one from which one can obtain - at least - the "post town", or which is the special code for "non-UK resident" or "parked and unattended".)
5. The quality of the postcodes that are being supplied is, in general, very good indeed. The almost 53,000 driver and casualty records for 1999 and 2000 that had been provided by Scottish Police Forces at the time of writing have been examined, and compared electronically with the Postcode Index file. In only 0.4% of cases had a "postcode" that is obviously wrong been supplied - e.g. a case where it appeared to have been mis-typed in such a way that the computer program could not identify even the post town or where the postcode field contained something which could not be a postcode (like "0" or "AAA"). The overwhelming majority of the "uninformative" postcodes were cases where either the code for "not known" had been used or the field had been left blank, presumably because a postcode (or part of a postcode) was not readily available for the person's home address. There did not appear to be many incomplete postcodes (e.g. where, say, only "EH10 " or "G1 " had been entered in the field), or slightly erroneous postcodes (e.g. where an error towards the end of the field, such as mistyping a letter as a digit or vice versa, meant that it did not match with the Postcode Index): in only 5% of cases was it impossible to verify the postcode, while being able to determine one (or more) of the following: the postcode sector, the postal district and the post town.
6. It now appears that almost all the main technical problems affecting the supply of postcode data have been overcome. Taking the averages of the latest three months for which returns had been received at the time of writing, the percentages of postcodes which are "informative" provided by the different Police Forces are as follows (in ascending order):
- Driver postcodes: 49, 65, 79, 79, 86, 89, 93
- Casualty postcodes: 56, 62, 80, 87, 87, 88, 93, 93
It will be noted that only seven figures are given for driver postcodes. This is because a (presumed) computer problem has suddenly affected the supply of driver postcodes for one Police Force. (An average of 76% of that Force's driver postcodes had been informative in the most recent three unaffected months. The supply of casualty postcodes has not been affected: in the latest three months, an average of 88% of its casualty postcodes have been "informative" ones.)
7. These figures show that three Police Forces are providing informative postcode details for more than 85% of drivers, and five Police Forces are doing so for more than 85% of casualties, which is very encouraging. However, there are two Police Forces which have particularly low percentages.
8. Of course, there will always be some drivers and casualties for whom it will never be possible for Police Forces to obtain addresses (e.g. "hit and run" drivers), and for whom postcode information can never be provided. This might not be a problem, if such drivers and casualties represent a small percentage of the total, and if they are likely to be (in terms of their home addresses) a typical cross-section of all driver and casualties. However, the results could be affected by a significant bias if the postcode information is unavailable for many drivers and casualties, and if these people are (in terms of their home addresses) not "typical" people. For example, suppose that postcode details were readily available for the vast majority of drivers and casualties who lived locally, and that postcode details were not readily available for those who lived further away - so that, perhaps, "not known" was recorded for most drivers and casualties from outwith Scotland (instead of, say, the relevant part postcode or non-UK resident code). Then, analysis of the postcode data would be misleading, because it would suggest that far too small a percentage of the drivers and casualties involved in accidents were from outwith the local area. Clearly, the higher the percentage of "not known" (or blank) postcodes, the greater the concern about the possibility of such biases.
9. Members of the Group are invited to comment upon the likelihood of the data being affected by such biases, and to suggest what might be done in future. For example
- could Police Force representatives say something about the kinds of drivers and casualties for whom postcodes are "not known" (or blank)?
- do Police Force representatives think that, in practice, a disproportionate number of the "not known" cases are not "typical" in some way (e.g. non-local) - and, if so, what might be the scale of the problem?
- how might Police Forces reduce the percentages of the driver and casualty records in their forthcoming "STATS 19" returns for which postcode details are not known? Comments would be particularly welcomed from representatives of the Police Forces which have very low percentages of "not known" postcodes at present, as others may be able to benefit from their experience and expertise.
- might Members of the Group feel that it would be helpful to agree targets for the minimum percentages of drivers and casualties for whom postcode information should be provided in future returns?
Research on Tourist Road Accidents in Rural Scotland: Project Details
10. The Scottish Executive is to commission a project in early summer on Tourist Road Accidents in Rural Scotland. For some time it has been the perception of some local authorities and residents in rural 'tourist' Scotland that a significant proportion of road accidents in these areas involve tourists (especially foreign drivers) visiting the area, who are unfamiliar with the layout of the roads and the nature of the traffic in that location.
11. So far, no research has taken place at a national level to determine whether there is a significant road accident problem posed by tourists visiting rural Scotland, or to explore the circumstances in which these accidents occur. This research will therefore be commissioned to investigate the number and type of road accidents involving non-local drivers in tourist areas in rural Scotland in order to inform the Scottish Road Safety Campaign on the nature and extent of the problem.
12. Specific objectives of the research are set out below:
- To set accidents in tourist areas in the context of general accidents in Scotland as a whole and establish whether there is a particular problem involving visitors to the area
- To determine whether there is a particular problem with foreign drivers or whether accidents are equally likely to involve UK visitors to the area,
- To establish whether one type of accident, or pattern of accidents, predominates more than others and on what type of road these tend to occur
- To ascertain the cause of tourist accidents, and whether this is due to factors such as unfamiliarity with the road layout or type, distraction or failure to understand road signing
- To provide recommendations on road safety strategies to address any problems found to exist
Research Stage 1: June to September 2000
13. Contractors will be asked to focus on the three case study tourist areas of Argyll & Bute, Highland and Aberdeenshire to investigate the nature and circumstances of accidents in these areas for the summer months of June, July and August. Clear language descriptions of accidents held by the relevant Police Forces in these areas (as part of the STATS 19 recording system) will be examined. Contractors will be expected to visit the Police Force offices in question in order to collate and extract this data which may be held in paper or electronic formats.
14. The three Police Forces in question have been written to, explaining the nature of the research and seeking co-operation for access to the data. Police in these areas have also been asked to be particularly alert to the need to collect accurate postcode or foreign driver information data over the coming summer months.
Research Stage 2: October 2000 to January 2001
15. The second stage of the research will involve secondary statistical analysis of STATS 19 data for the summer months of June to August 2000 inclusive for all of Scotland so that accidents in rural 'tourist' areas can be set in the general Scottish context . This data will be made available to the contractors by the Scottish Executive. Central Research Unit has written to ACPOS asking that all police forces in Scotland take special care over the coming months to collect as accurate information as possible on the postcode or 'non-UK' status of the driver, and hopefully the quality of such data will have improved from ast year, by the time of analysis for Scotland as a whole is carried out.
16. Members of the Group are invited to note that this research project is being commissioned, and that Police Forces across Scotland have been asked to take special care to collect as accurate information as possible on the postcode or 'non-UK' status of the driver.