LGRAS (02) 3 - THE SUBMISSION AND THE USE OF THE "STATS 19" DATA
1.1 This paper covers various points to do with the submission and the use of the STATS 19 returns for Scotland. Members of the Group are invited to comment on these matters at the meeting, and to contact the Scottish Executive Transport Statistics branch regarding any subsequent developments or difficulties.
1.2 In April, Alastair Douglas left Transport Statistics for a post in an Animal Health and Welfare policy branch. At the time of writing, we do not know when a replacement will arrive. Therefore, some aspects of the branch's work on road accident statistics will have to be delayed, or not done at all, until such time as Alastair's eventual successor is settled in post For example, being short-staffed, we may be unable to respond swiftly on matters raised by data suppliers and users. We regret any inconvenience that this may cause.
1.3 To facilitate communication in future, Members of the Group who are STATS 19 data suppliers are invited to provide Transport Statistics with their email addresses by sending an email to: firstname.lastname@example.org (unless they have already done so).
2. Submission of STATS 19 returns
2.1 At the LGRAS meeting on 24 April 1998, Members of the Group representing data suppliers agreed that every return (apart from the "December" return) should be made within 6 weeks of the end of the month to which the return relates. For example, the "January" return should be made by around 15th March, unless there were major unforeseen problems (such as those that arise from the conversion of computer systems). Because of the backlog which builds up over Christmas / New Year, it was agreed that the "December" return was an exception, and should be sent to the Scottish Executive by the end of February.
2.2 Because of the staff vacancy, we have not prepared the table that usually accompanies these reports, which shows when the Road Accident statistics returns which were due to be received in the period (broadly) since the previous meeting actually reached the Scottish Executive. Instead, we simply report on the current position (23 May 2002: a day or two before the papers for the meeting are copied and sent out).
2.3 At the time of writing, we should have received all the returns for March 2002 (on the basis of the "six weeks" agreement, they should all have been made by the middle of May 2002). The latest returns which we have received are for the following months:
December 2001 - one area (where a new Police system has "teething troubles")
January 2002 - one area
March 2002 - three areas
April 2002 - three areas
The position is much better than it was in previous years. The Police Forces which had large backlogs of overdue returns have caught up, for which we are grateful. Some Police Forces have been making regular returns within two weeks of the end of the month, for which they are congratulated . We appreciate the efforts made by all the data suppliers, which have resulted in a more timely submission of data. This will be increasingly important once DTLR starts to release quarterly statistics for Great Britain as a whole, at which point the Scottish Executive may also start to issue its own quarterly figures (see sections 5 and 6 of the paper).
2.4 Members of the Group representing those data suppliers are asked to ensure that henceforth they make their returns within the agreed timescale (that described in paragraph 3.1.)
3 Identifying inconsistencies between the grid co-ordinates of the location of the accident and the local authority code
3.1 It is well known that there are occasionally "large" errors in the grid co-ordinates allocated for road accidents (e.g. a few accidents have the grid co-ordinates of points which are well out at sea). There have also been cases where a local authority has found that accidents which occurred in one Council area have been given the local authority code of another Council area.
3.2 In order to reduce the number of such errors in its copy of the data, the Scottish Executive introduced a new procedure a few months ago. Now, the Scottish Executive Geographic Information Service (SEGIS) uses the grid co-ordinates shown for the accident:
(a) to check that the location reported for the accident is on land; and, if it is
(b) to determine in which local authority area the accident appears to have occurred (ifthe grid co-ordinates are correct).
3.3 Transport Statistics then produces a printout listing
(i) apparent errors in the grid co-ordinates - cases where the reported grid co-ordinates do not appear to fall within the land boundaries of any Scottish local authority; and
(ii) apparent inconsistencies between the grid co-ordinates and the local authority code that was allocated for the accident - cases where, to SEGIS, the grid co-ordinates appear to be for a point in another Council area.
This is sent to Police Forces, who investigate the cases listed and inform Transport Statistics of any corrections that may be required.
3.4 It should be noted that some of the apparent inconsistencies identified in these printouts do not represent errors. For example:
- some accidents which occur on the Forth Road Bridge have grid co-ordinates which, in terms the SEGIS system, do not fall within the boundaries of any local authority - so a query may be generated even though the grid co-ordinates are perfectly correct.
- in the STATS 19 system, grid co-ordinates are recorded only to within 10 metres (as only 5 digits are available for use to record each of the Easting and the Northing) - so an accident which occurs close to the boundary between two local authorities may have grid co-ordinates which are correctly recorded to the "nearest" 10 metres and a correct local authority code, but these may appear inconsistent to the SEGIS system (because the point with those exact grid co-ordinates happens to be on the other side of the boundary, and therefore, in the SEGIS system, has a different local authority code).
Unfortunately, because of the nature of the data, it is not possible to prevent the printouts from including such queries.
3.5 Members of the Group representing data suppliers are asked to bear in mind the fact that the printouts identify cases where it appears that one or both of the grid co-ordinates, or the local authority code, may be incorrect. However, the fact that it appears to the SEGIS system that the data are inconsistent does not mean that there has necessarily been an error - the circumstances may be such as those described above, in which case a "false alarm" may have been raised.
4 Procedures in cases in which local authorities decide that an accident has been assigned the wrong local authority code
4.1 It appears that, from time to time, a local authority may decide that an accident which has been given its local authority code actually occurred in a neighbouring local authority area. The "January 2001" newsletter of the Standing Committee on Road Accident Statistics (copies of which were distributed under cover of Alastair Douglas' letter of 30 January) included a short note on this, which stated that "….in some other cases individual authorities were removing 'out of area' accidents without notifying the appropriate neighbouring authority. In most cases, these alterations were not notified to the Police or DETR, which has resulted in some discrepancies between the figures published in Road Accidents Great Britain and those published locally."
4.2 The use of the SEGIS system to check the consistency of grid co-ordinates and local authority codes cannot identify cases where an accident is "relocated" by a local authority, on the basis of the information which is included in the free-text description of the accident location (as that is not provided to the Scottish Executive). Members of the Group are therefore invited to consider whether the number of such accidents is likely to be "significant" in any circumstances (e.g for those who are looking at the accident statistics for particular routes), whether any further investigation is required, and whether there may be a need for arrangements for local authorities to report to their neighbours, Police Forces and the Scottish Executive details of the accidents which, they believe, should be allocated to other local authorities.
5 The provision of statistical information when police officers attend accidents which occurred outwith their Police Force area
5.1 It appears that, from time to time, officers from one Police Force will attend an accident which occurred in the area of another Police Force. This can happen if, for example, most of the road between two places is within the area of one Police Force, but it happens to cross and re-cross a boundary, with the result that a small stretch (which is not at one or other end of the road) is within another Police Force's area.
5.2 A few such cases were found in the data submitted to the Scottish Executive, when the checks using the SEGIS system (described in section 3) identified inconsistencies between the grid co-ordinates and the local authority code. In these cases, it was found that the grid co-ordinates were correct, and that the accident had wrongly been assigned the code for a local authority which was within the reporting Police Force's area. There may, of course, be other cases of "out of area" accidents which have not been identified, for example because they have been assigned grid co-ordinates which are within the reporting Police Force's area, and therefore appear consistent with the local authority code that they have been given.
5.3 Such cases could cause a disproportionate amount of administrative difficulty. The Police Force code cannot be changed to the code for the Police Force for the area in which the accident occurred, because the latter is not the Police Force which reported the accident to the Executive. Changing the local authority code to the correct value would create an inconsistency between the Police Force code and the local authority code. Indeed, the relevant local authority might not have been received any details of the accident, since the reporting Police Force's system records the accident as having occurred in a different local authority.
5.4 Members of the Group are invited to consider whether the number of such accidents is likely to be "significant" in any circumstances (e.g for those who are looking at the accident statistics for particular routes), whether any further investigation is required, and whether there may be a need to consider any changes to the current arrangements which Police Force have for reporting accidents to the Scottish Executive, the relevant local authorities and to other Police Forces.
6 The procedures for identifying accidents which occurred on trunk roads
6.1 A local authority representative recently queried why certain accidents had not been allocated to the trunk road network in the Scottish Executive's database. It may therefore be helpful to describe and discuss the procedure which the Executive's contractor, WDM, uses to identify accidents which occurred on trunk roads.
6.2 The process has a number of stages. It should be noted that the only information that WDM uses is that which is contained in the STATS 19 returns: WDM has the road class and number, and the grid co-ordinates reported for the accident location. Therefore, WDM may well mis-allocate some accidents for which the reported road class / number or the grid coordinates are wrong (e.g. first road class / number were wrongly shown as "unclassified"). WDM does not receive the free-text descriptions of the accident location and the accident "story" (as they are not part of the STATS 19 returns), so WDM cannot use them to identify cases where the wrong road class / number and/or grid co-ordinates have been reported. WDM carries out some checks, which are described below. In cases where WDM changes or queries the details that were supplied, its system produces a "location change / query" print (along the lines of the illustrative example that was enclosed with paper LGRAS  8), which is sent (via Transport Statistics) to the relevant data supplier, who can use the description of the accident location, and any other information about the accident, to answer the query.
6.3 Accidents for which the first road class/number are those of a trunk road:
- these are automatically "fitted" to the trunk road network, if the grid co-ordinates appear to be right (it must be remembered that parts of some roads are trunk, and other parts are not - for example, some accidents on the A7 occur on the parts which are trunk, and some do not);
- if the co-ordinates do not appear right, the details are checked manually (eg in case the Northing and the Easting have been transposed). WDM then produces a "location change / query" print, if appropriate .
6.4 Accidents which do not have the first road class/number of a trunk road are put into a separate database, and:
- checked for possible "typos"
- plotted on a map - those which fit, or come close to, the trunk road network are raised as queries
- in both cases, WDM will produce a "location change / query" print, if appropriate,
6.5 WDM has now introduced a further check, based on the second road class/number. If the second road class/number is that of a trunk road, the accident is examined manually, on an individual case basis. In the first month in which this check was used, it identified only one more apparent trunk road accident: a case in which the first road/class was "unclassified" and the second road/class was that of a trunk road. As the road type was "dual carriageway" and it was clear from the map that the unclassified road was not a dual carriageway, WDM concluded that the two road class/numbers had been reported "the wrong way round" in STATS 19.
6.6 Members of the Group are invited to ask questions about, and comment upon, this matter. For example, do they feel that there has been any mis-allocation of accidents to the trunk road network? If so, are the numbers that have been misallocated significant? Are there ways in which information used in the procedures, or the procedures themselves, could be improved? For example, should the Scottish Executive send each local authority a printout showing all the accidents in its area which were allocated to the trunk road network? If so, how frequently should this be done? Improving the quality of the data would involve additional work for those who would scrutinise the accident details, in order to raise and respond to queries: do Members of the Group feel that the benefits would be likely to outweigh the costs?
7 Publication of the figures for Scotland
7.1 Provisional totals for 2001 will be published on Thursday 13th June in Key 2001 Road Accident Statistics. Copies of this bulletin will be sent to Members of the Group at the time (please contact Andrew Knight if you have not received a copy by, say, Monday 24 June).
7.2 The "final" detailed figures for 2001 should be published before the end of November in Road Accidents Scotland 2001. The precise timing is uncertain, and will depend upon the effects of the current staff vacancy. Members of the Group will receive their copies shortly afterwards.
7.3 Members of the Group are invited to note these recent and forthcoming publications.