Scottish Government

7. The DfT survey of Public Service Vehicle Operators

7.1 General information

The basis for the statistics in Tables 1.1 to 10.1 is the annual returns which a sample of Public Service Vehicle operators across GB make to the Department for Transport (DfT). Further information is contained in the latest editions of DfT publications Transport Statistics GB and Public Transport Statistics Bulletin GB. Enquiries regarding the statistics should be made to Chris Mottau of DfT (tel: 0207 944 3076). From 2004-05 onwards, the survey has collected less information on receipts from non-local services (such as coach work and private hire) and put more emphasis on the income from local services of large bus operators. As smaller operators who do non-local work are less well-represented, reliable estimates of non-local receipts are not available for Scotland or for other parts of GB.

The sample is stratified by size of operator, based on the numbers of vehicles that they are licensed to run (their numbers of licence discs). With effect from the 1997-98 survey, it includes all operators with 21+ vehicles, and a random sample of smaller operators (who are selected with probabilities depending upon their numbers of discs and, from 2005-06, whether they provide local bus services).

The sampling frame is sorted geographically by postcode running from the South West ( TR -Truro) to the North East ( ZE - Shetland), with the aim of ensuring an even coverage of GB. However, it is not possible to ensure that the sample for an area within Scotland is representative: the "luck of the draw" could lead to the sample for a particular area being unrepresentative of the operators in that area. There is an additional problem if there are few operators in an area: for example, suppose that it has only one operator in a size-band for which the sampling fraction is 25%. On average, that operator should be included in the sample in a quarter of the years - so, in most years, that area will appear not to have an operator in that size-band.

In order to improve the estimates of local bus passenger numbers, DfT changed the survey's methodology with effect from 2005-06:

• the probability of selection is now greater for the small operators which provide local bus services than for those that do not; and
• "proxy" data are now used for all local bus operators that are not in the sample or do not respond for any reason. The proxy values are taken from data for the same operator in an earlier year or for an operator of similar size (based on vehicle-kilometre information available from DfT's Bus Service Operators Grant database). The effects of any errors in these estimates are likely to be small since only 2% of passenger journeys are based on such proxy data.

As a result, with effect from 2005-06, "grossing-up" of data (see section 7.3) is not required for operators who provide local bus services.

The results for non-local operators, and for operators of local bus services for the years prior to 2005-06, will be less precise, because the data for these operators have to be grossed-up to compensate for those who were not sampled or who failed to respond.

The figures for Scotland for non-local services are derived from the returns made by the operators who are based in Scotland, and so will include services into England which are run by operators based in Scotland, and will exclude services into Scotland which are run by operators based in England. ( NB: a large group, such as Stagecoach, is not treated as a single operator: there will be a separate statistical return for each of its subsidiary companies.)

Sampling fractions and 2005/06 distributions by numbers of discs( GB figures)

N.B. The sampling fractions and size-bands used have changed over the years.

* See section 7.3. The actual "grossing-up" factors for each size band depend on the level of response. The factors shown here assume 100% response rates, but the factors actually used are higher, because of non-response.
** Small operators, in particular, enter or leave the industry throughout the year.

7.2 Estimation of certain breakdowns, in cases where the survey does not collect information for smaller operators

In order to reduce the burden on respondents to the survey, the smaller operators receive a simplified questionnaire, which does not ask for certain pieces of information. For example, smaller operators supply only an overall total figure for their income from local services: unlike the larger operators, they do not provide a breakdown between "passenger receipts", "public transport support", "concessionary fare reimbursement", "contract payments" and "Bus Service Operators Grant".

In order to produce overall totals for the different types of income from local services, DfT must estimate the breakdown of local service income between these headings for the smaller operators. DfT does this by assuming that, for each small operator which responds to the survey, the percentage breakdown of income between these headings is the same as the overall Scotland percentage breakdown of such income for all the large operators taken together. Thus, if "passenger receipts" accounted for (say) 80% of the total local service income of all large operators in Scotland, DfT would assume (for the purposes of estimating the overall totals) that passenger receipts accounted for 80% of the local service income of each of the smaller operators who responded to the survey.

The survey obtains, from the larger operators, information about their expenditure (operating expenditure on local services and on other services, administration expenditure and depreciation). It does not ask the smaller operators to provide this, but the effect of this on the calculation of costs per vehicle kilometre and per passenger journey is thought to be small.

7.3 Estimation of totals for all smaller operators, by "grossing-up" the returns from the smaller operators

NB: with effect from 2005-06 (see section 7.1), "proxy" data are used for the small local operators which were not sampled (or did not respond), so only the data for the smaller non-local operators are "grossed-up".

The figures for each smaller operator (including those income breakdown figures which DfT estimated - see above) are "grossed-up" using a grossing-up factor which is the inverse of the achieved sampling fraction, and therefore makes an allowance for any non-responding small operators (as the grossing-up factors are based upon the numbers of responses obtained for each fleet size group). For example, if roughly 5% of the operators in a particular size-band respond to the survey, the grossing-up factor for them will be about 20. (While the aim may be to sample 1-in-20 of the operators in a particular size-band, the survey is unlikely to obtain returns from exactly 1 in 20 of them: for example, if there were 250 such operators, and 12 of them were chosen for the survey and made returns, the grossing-up factor would be 250 divided by 12, which is about 20.8).

7.4 Estimation of figures for areas within Scotland

The larger operators (those with over 1 million vehicle-miles run on local services) are asked to split their local service passenger numbers, vehicle miles and income of various kinds between the local government areas in which they operate. Up to and including the 1996-97 returns, the breakdown was requested in terms of the former Regions and Island Areas; from the 1997-98 returns it is in terms of the present Councils. In order to provide the longest-possible time-series, the present Councils have been grouped together on the basis of the former Regions to produce Tables 9.1 and 9.2. Regional Transport Partnership area figures (also based on groupings of Councils) are shown in Tables 9.3 and 9.4; more information on the RTP areas is given in section 6.1.

The methods that operators use to split their totals between areas will depend upon the kind of information that is available from their businesses' records, and may vary from operator to operator: some may have administrative systems that provide breakdowns of passenger numbers etc by area that are more detailed than those of other operators. There is no recommended method: all that the instructions for the completion of the return ask is that "if exact figures are not available, please give the approximate proportion or percentage in each [area]". In some cases, therefore, an operator might only be able to provide a rough estimate of the breakdown (e.g. 60% in area A, 30% in area B, 10% in area C). It may be that, from time to time, an operator changes the way in which it estimates or calculates the split of its figures between the areas which it serves - if so, this may cause a discontinuity in the time-series for those areas.

The smaller operators are not asked to provide such a breakdown. Instead, DfT assumes that all a small operator's activity is in the area in which the operator's office is situated.

7.5 DfT's rules on the possible disclosure of figures for an area

When the data are collected from Scottish operators, an undertaking is given that (unless they give written permission) access to their figures will be restricted to the DfT and Scottish Executive staff who deal with public service vehicle statistics. Therefore, the need to protect the confidentiality of the information which is provided in the statistical returns may prevent the disclosure of the totals for an area without the permission of the larger operators in that area. For example, if the two largest operators in an area together account for a very large proportion of the total patronage, one of them might be able to deduce the approximate size of the other's business from any totals that were published for that area. In such a case, therefore, one would need the permission of both large operators before one could disclose the totals for that area. DfT uses the following rule to determine whether or not the totals for an area may be disclosed without the permission of the larger operators in the area:

a total for an area is NOT disclosable (without the permission of the larger operators in the area) if

either i. the largest operator accounts for over 60% of the total for the area

or ii. the second largest operator accounts for over 60% of the rest

These rules mean that figures for several parts of Scotland cannot be disclosed without the permission of the larger operators in those areas. Therefore, as the collector of the statistics, DfT kindly wrote to the largest operators in Scotland, asking if they would agree to the figures from their statistical returns being used to produce totals for areas within Scotland which the Scottish Executive would publish. We are very grateful to all the major operators who agreed that the Scottish Executive could use their figures in this way.

7.6 The uncertainties about the estimates for parts of Scotland

It must be remembered that the survey was not designed to produce reliable figures for areas within Scotland (such as the former regions).

There are some large year-to-year changes in the estimates for some areas, and occasionally one year's figure stands out as unusually high or unusually low. This is thought likely to be largely due to operators having difficulty assigning bus passenger boardings and vehicle kilometres to particular local authority areas (see section 7.4).

In addition, for 2004-05 and earlier years, the estimates for any areas which have only a few operators could be subject to considerable sampling errors, if those operators are small operators, since the "luck of the draw" could include most of the operators in the sample in one year (leading to the grossed-up estimates being far too high) and few or none in another year (leading to the grossed-up estimates being too low). Indeed, unreliability may be inevitable when an area has few operators. For example, suppose that an area has only one operator in a size-band for which the sampling fraction is 25%. In the years in which that operator responds to the survey, its figures will be multiplied by a grossing-up factor of (at least) 4, so the survey's estimate for the area will be too high (all else being equal). In the years in which that operator is not in the sample (or does not respond), it will make no contribution to the survey's estimate for the area, which will be too low (all else being equal). Such factors could have a considerable effect on the figures for 2004-05 and earlier years. However, they should not affect the results for 2005-06 onwards, as the survey now uses "proxy" data for the local bus operators who are not in the sample (or who do not respond). Whether the survey will provide reliable local bus passenger numbers for areas within Scotland for 2005-06 onwards will depend upon how accurately the large operators can split their passenger numbers between the different areas that they serve.

It follows that some of the estimates that are provided in Tables 9.1 to 9.4 may be subject to considerable uncertainty. For example, an area could appear to have a particularly large increase in its estimate in one year because:

one of the larger companies which operates in that area changed the way in which it split its totals between the areas in which it operates, with the result that a much greater percentage of its total was counted against that area;

or a smaller company, which operates in that area but is based in another area, has now become large enough to be required to split its totals between the areas in which it operates, and therefore provides figures for this area for the first time ever;

or in that year (if it was 2004-05 or earlier), the survey obtained returns from disproportionately more of the operators who are based in that area;

or in the previous year (if that was 2004-05 or earlier), the survey obtained returns from disproportionately few of the operators who are based in that area;

or a combination of some of the above factors, and other factors, produces that result.

The estimates for different parts of Scotland provided in Tables 9.1 to 9.4 are therefore a general guide to the way in which the levels of patronage and service have changed in each of these areas over the past 10-or-so years. The "hiccups" in the series are a reminder that the estimates may be affected by sampling quirks etc, and so are not necessarily reliable. In order to see better the overall trends in local bus service provision and patronage in some areas, users of the statistics could take (say) three-year moving averages of the estimates for those areas, as moving averages should be less affected by sampling and estimation problems.