Introduction
As part of our continuous review of road traffic methodology, the Department for Transport’s (DfT) Road Traffic Statistics Team are undertaking a deep dive into minor road traffic methodology. The scope of this review has been split into two areas:
- a review of historic minor road traffic estimates
- an investigation into methods for improving the robustness of future annual minor road traffic estimates
This report summarises the outcome of the historic minor road traffic estimate review, the conclusions of which have been implemented in the latest 2021 Road Traffic statistics publication.
The report also includes further details of planned areas for further investigation and development on future annual minor road traffic estimates.
We are keen to receive user feedback on the issues covered in this document. This can be given via the Road Traffic Statistics Team inbox.
Background
The DfT publishes annual and quarterly estimates of road traffic on Great Britain’s roads. In 2020 the department revised its minor road traffic estimates for 2010 to 2019 as a result of the latest minor roads benchmarking exercise, which is conducted every decade. A summary of how minor road traffic estimates are typically produced is available in the annex section of this document.
The methodological approach used for the 2019 minor road traffic benchmarking exercise was developed and verified in conjunction with independent statistical methodologists from the Office of National Statistics. However, the adjustment applied to minor road traffic as a result of the 2019 benchmarking exercise and associated data collection was higher than that of previous benchmarking exercises.
As part of the subsequent minor road traffic estimates review, the DfT’s Road Traffic Statistics Team has carried out further exploration of previous benchmarking exercises and alternative annual minor road traffic estimation methods to investigate any improvements that could be made to the historical series whilst ensuring the consistency of the time series.
Summary
This technical report details the areas of investigation into historic minor road traffic estimates. These investigations have resulted in revised estimates of historic minor road traffic.
The overall impact of these revisions is displayed in chart 1. Previously published minor road traffic estimates showed a 26% increase in minor road traffic in Great Britain between 2009 and 2019. After applying the revisions detailed in this report, newly published minor road traffic estimates show a 10% increase in minor road traffic over the same period. The total estimate for 2009 has been reduced by less than 1% and the estimate for 2019 has been reduced by 13%.
Chart 1: Estimated minor road traffic (billion vehicle miles) in Great Britain before and after applying historic revisions, 1993 to 2020
Areas of Exploration
This section will cover each area investigated as part of the review of historic minor road traffic estimates, detailing the exploration which took place and the impact of any resulting amendments to minor road traffic estimates.
The use of an additional data source, GPS data, in the minor road traffic benchmarking methodology was important in two of the areas of exploration.
GPS data
The department purchases GPS data from commercial providers. The data collection includes a sample of over 100,000 vehicles with a fixed GPS device, often fitted for security, insurance, or fleet monitoring purposes. The resulting dataset guarantees daily coverage on at least 95% of the Strategic and Local ‘A’ road network. The contractual coverage is not the same for the minor road network, but there is wide coverage over a longer time period, with coverage of around 75% of minor roads in England and Wales over a calendar year. The ‘GPS dataset’ provides an observation count of how many times the sample of vehicles with a fixed GPS device travelled along each stretch over the course of a given year. For the purposes of this review, the dataset has been limited to only minor roads. These vehicle observation counts provide an indication of how busy minor road links were, relative to each other.
Throughout this report this data will be referred to as the ‘GPS dataset’. Further information regarding this dataset, including limitations, is detailed in the annex of this document.
1. Comparing the sample of minor road locations used for the previous two benchmarking exercises
The sample of minor road locations used in 2009 differs from that used in 2019. This is by design and is explained in the road traffic estimate methodology note in the annex of this document. Part of the review investigated if there were any fundamental differences between these two benchmark samples that would have led to the higher benchmark adjustment seen in 2019.
Using the GPS dataset, all minor road links in England and Wales were assigned as being either a ‘high’ flow link or a ‘low’ flow link. Based on GPS observations, a link was assigned as ‘high’ flow if it had more than the median number of observations in the given year; otherwise the link was assigned as ‘low’ flow. Around a quarter of minor road links did not have an observation in the GPS dataset so a third ‘Not known’ group was added for these. The distribution of ‘high’ and ‘low’ flow sites assigned to each stratum was analysed for both benchmark exercises. The findings showed that the sample selected for the 2009 benchmark had a higher proportion of ‘low’ flow sites than the sample selected for the 2019 benchmark. As this inconsistency between the chosen samples had not been controlled for in the calculation of minor road traffic estimates, it may have contributed to the difference between the 2009 and 2019 benchmark traffic estimates.
Table 1: Of the minor road count points which were assigned a flow grouping, table 1 gives the proportion of count points categorised as being on a low or high flow road link
| Benchmark year | Low flow | High flow |
|---|---|---|
| 2009 | 44% | 56% |
| 2019 | 39% | 61% |
Outcome: To control for this difference in the distribution of flows between the two benchmark samples, ‘flow group’ was added as an extra stratification factor for the recalculation of both the 2009 and 2019 minor road benchmarking exercises. ‘B’ roads and cul-de-sacs were not assigned a flow group as the introduction of an additional stratification factor would have led to insufficient sample sizes.
2. Consistency of cul-de-sac methodology
The methodology used when calculating minor road traffic estimates from the 2009 and 2019 benchmarks were mostly consistent. Where differences exist, the 2019 methodology is considered more robust; it is the most up to date and the most recently verified in conjunction with independent statistical methodologists from the Office of National Statistics.
However, the notable difference in the 2019 benchmark methodology was the identification of cul-de-sac roads. The pattern of traffic on cul-de-sacs is different from that on other roads, with vehicles being less likely to drive the full length of the road. Therefore, in the 2019 benchmark traffic on cul-de-sacs was halved in the methodology, so as not to be over-represented in the final traffic estimates. This was not originally taken into account when calculating traffic estimates from the 2009 benchmark.
Outcome: The 2009 benchmark estimates were recalculated using the same cul-de-sac methodology as the 2019 benchmark exercise.
3. Stratification of data from London
It has previously been identified that trends in published road traffic estimates in London over the past 20 years have differed from those in other regions. In both the 2009 and 2019 benchmarking exercises, ‘London’ was treated as a single region for stratification purposes. From analyses of major road traffic estimates, it is evident that traffic patterns differ between Inner London and Outer London (see chart 2). Therefore, this review analysed the effect of including Inner and Outer London as two separate levels in the minor roads benchmark regional stratification.
Chart 2: Change in major road traffic estimates in London, 2009 to 2019
Outcome: The 2009 and 2019 benchmark estimates were recalculated with Inner and Outer London included as two regions. The impact of this change, combined with the introduction of the flow group stratification, produced minor road trends in London that are similar to those seen on the major roads. It led to an overall increase in the minor road traffic estimate for London in 2009 and a decrease in 2019.
4. Urban and rural road classification for the 2009 benchmark
The 2009 benchmark included stratification of ‘B’ roads and ‘Classified unnumbered and unclassified roads’, which is the same as that used in the 2019 benchmark stratification. Both benchmarks also included an urban or rural identification to the sample, but investigations into the 2009 data determined that the urban or rural identification did not match the definition currently used for road traffic estimates, which is based on location within an urban area with a population of size 10,000 or more.
Outcome: It was not possible to include an urban or rural stratification in the recalculation of the 2009 benchmark.
5. Consistency of annual methodology
Between benchmarks, annual minor road traffic estimates are calculated by multiplying the previous years’ minor road traffic estimates by the change in traffic and the change in road lengths compared to the previous year. The change in traffic since the previous year is estimated from traffic counts undertaken each year at a representative sample of minor road sites.
Analysis of minor road traffic estimates for the years 2000 to 2019 identified some inconsistencies in the application of the annual methodology in some years, which have been corrected as part of this review.
| Applicable years | Detail of revision |
|---|---|
| 2000 to 2009 | The original 2009 benchmark adjustment for motor vehicle traffic was also applied to pedal cycle traffic. As pedal cycle trends can vary from those of motor vehicle traffic, no benchmark adjustments have been applied to pedal cycle estimates in the revised series. |
| 2004 | The 2004 road traffic estimates were originally calculated using 2003 road lengths. The correct road lengths have applied in the revised series. |
| 2004 to 2019 | In the original series, road traffic estimates were missing for small lengths of road for a few road categories in a few local authorities. These have been included in the revised series. |
| 2012 | The original 2012 road traffic estimates were calculated using the change in traffic between 2010 and 2011. The revised road traffic estimates have been calculated using the change in traffic between 2011 and 2012. |
| 2013 | The original 2013 road traffic estimates were calculated using a provisional version of the 2012 traffic estimates, rather than the final published 2012 minor road traffic estimates. The published 2012 estimates have been used in the revised series. |
6. Validity of historic data points
Analysis of the GPS dataset showed it has clear utility as an independent data source that can be used to validate historic road traffic counts. As part of the review, the GPS dataset was used to identify significant deviations between observations from the GPS dataset and calculated annual average daily flow (AADF) estimates from the 2009 and 2019 benchmark samples.
Outcome: This resulted in 46 count points being removed from the 2009 benchmark exercise and a further 9 count points having their road classification changed but retained in the sample. Note that this is out of a total of 10,483 count points from the original 2009 benchmark exercise. The most common reason for these removals and reclassifications was that a manual count had been conducted on a high flow ‘A’ or ‘B’ road which was parallel to the Classified unnumbered or Unclassified road that was sampled and where the count was meant to have taken place.
For example, figure 1 shows count point 948776, which was removed from the 2009 benchmark. The count point is on a relatively low flow Unclassified road, which runs parallel to the A4123. Both the Unclassified road and the ‘A’ road are named ‘Birmingham New Road’. The AADF calculated at this count point for the 2009 minor roads benchmark was 23,093. It is reasonable to assume that the manual count used to calculate this AADF took place on the parallel ‘A’ road.
Count Point 948776 on a map showing that the count point is on an Unclassified road, parallel to the much busier A4123.
Contains OS data © Crown Copyright and database right 2022 Contains data from OS Zoomstack
Impact on historic minor road estimates 2000 to 2020
The improvements set out in the previous section “Areas of Exploration” were implemented to the historic minor road traffic estimates for the years 2000 to 2020. As a result, historic minor road traffic estimates have been revised and are lower than those previously published. The 2009 minor road traffic estimate for Great Britain has decreased by 0.8% and the 2019 estimate has decreased by 13.2%.
Table 2: Estimated minor road traffic (billion vehicle miles) for Great Britain before and after the minor roads review, 2000 to 2020
| Year | Before review | After review | Revision to estimate |
|---|---|---|---|
| 2000 | 103.2 | 103.1 | -0.1% |
| 2001 | 103.6 | 103.4 | -0.2% |
| 2002 | 107.2 | 106.8 | -0.3% |
| 2003 | 107.3 | 106.9 | -0.3% |
| 2004 | 107.7 | 106.4 | -1.1% |
| 2005 | 107.9 | 106.9 | -1.0% |
| 2006 | 109.1 | 108.1 | -0.9% |
| 2007 | 111.8 | 111.1 | -0.7% |
| 2008 | 110.3 | 109.8 | -0.5% |
| 2009 | 108.1 | 107.3 | -0.8% |
| 2010 | 108.4 | 105.8 | -2.4% |
| 2011 | 109.4 | 105.5 | -3.6% |
| 2012 | 110.8 | 106.4 | -4.0% |
| 2013 | 112.7 | 106.6 | -5.4% |
| 2014 | 118.9 | 111.2 | -6.5% |
| 2015 | 122.1 | 112.9 | -7.5% |
| 2016 | 125.6 | 115.4 | -8.2% |
| 2017 | 130.1 | 117.4 | -9.7% |
| 2018 | 131.9 | 116.6 | -11.6% |
| 2019 | 135.8 | 117.9 | -13.2% |
| 2020 | 112.4 | 97.8 | -13.0% |
Chart 3: Estimated minor road traffic (billion vehicle miles) in Great Britain before and after applying historic revisions, 1993 to 2020
The level of revision to the previously published minor road traffic estimates varied by region, and over time. Within England, the level of revision to the 2019 estimates varied between 0.3% in the South West and 25.3% in the North West.
The main reason for the regional differences in revisions is the introduction of the stratification by flow group. For example, the proportion of sampled count points in the ‘high’ and ‘low’ flow groups in the South West barely changed between 2009 and 2019, so this change in methodology did not result in a large revision to the 2009 and 2019 minor road traffic estimates. The proportion of sampled count points in the North West which were ‘high’ flow increased by 17 percentage points between 2009 and 2019, leading to a larger revision to the 2019 estimate.
No region or country had an increase to their 2019 minor road traffic estimate as part of the revisions. Revisions to the 2009 traffic estimates were less significant, apart from in London with an increase of 22%; this was due to London having been split into two stratification levels for Inner and Outer London.
Table 3: Estimated minor road traffic (billion vehicle miles) by region before and after the minor roads review, 2009 and 2019
| Region / Country | 2009 estimate before review (bvm) | 2009 estimate after review (bvm) | Revision to 2009 estimate | 2019 estimate before review (bvm) | 2019 estimate after review (bvm) | Revision to 2019 estimate |
|---|---|---|---|---|---|---|
| East Midlands | 8.7 | 9.1 | 4.1% | 10.7 | 10.4 | -3.5% |
| East of England | 13 | 12.7 | -1.9% | 15.3 | 13.1 | -14.5% |
| London | 6.6 | 8 | 21.6% | 10.4 | 8 | –22.7% |
| North East | 4.5 | 4.2 | -6.2% | 5.8 | 4.8 | -17.4% |
| North West | 10.9 | 10.4 | -4.6% | 15.4 | 11.5 | -25.3% |
| South East | 17.1 | 15.5 | -9.4% | 19.7 | 17 | -14.1% |
| South West | 11.8 | 12.6 | 7.2% | 14.5 | 14.5 | -0.3% |
| West Midlands | 10.9 | 10.8 | -1.0% | 13.2 | 11.6 | -12.6% |
| Yorkshire and The Humber | 9.1 | 9.2 | 0.4% | 12.9 | 10.2 | -21.1% |
| England | 92.6 | 92.6 | -0.1% | 118.0 | 100.9 | -14.5% |
| Scotland | 9.3 | 8.9 | -4.3% | 10.1 | 10.1 | 0.0% |
| Wales | 6.1 | 5.8 | -5.6% | 7.8 | 6.9 | -11.2% |
| Great Britain | 108.1 | 107.3 | -0.8% | 135.8 | 117.9 | -13.2% |
Note that Scotland was not included in the 2009 benchmark exercise. The available GPS dataset, which is the basis of the most significant methodological changes introduced in this review to the 2009 and 2019 benchmark methodologies, only holds data for roads in England and Wales. Therefore, the revisions to minor road traffic estimates for Scotland are only based on the improvements detailed in the ‘Consistency of annual methodology’ section of this document.
Further Areas of Investigation
This update concludes the review into historic minor road traffic estimates. The next stage of the review will be to investigate options for improving the robustness of future annual minor road traffic estimates.
1. Additional data collection
During 2021, the department trialled an expanded scope for data collection of minor road traffic data. These trials were: increasing the frequency of counts by repeating manual traffic counts at a sample of sites during 2021; conducting traffic counts for a longer duration than 12 hours; and conducting counts outside of the usual counting season of May to October. Analysis of this data will consider if regularly performing these additional counts on a wider scale will help to improve the robustness of minor road traffic estimates.
2. Automatic Traffic Counter (ATC) data from Local Authorities
ATC data plays an essential role in the calculation of all road traffic estimates, including for minor roads. Data collected from the DfT’s network of around 300 ATCs are used in the production of annual average daily flow (AADF) estimates for each count point, which in turn are used in the calculation of minor road traffic estimates. See the ‘annual road traffic estimates: methodology note’ on the road traffic statistics information page for detail.
Many local authorities operate their own ATCs for local analysis independently of those managed by the DfT. The department is currently engaging with local authorities regarding the shared use of ATC data to improve the richness of datasets. Investigations will be conducted into the quality and suitability of ATC data from various local authorities and if it is feasible for their use to supplement the ATC data already collected by the DfT. This could improve the robustness of data summarised by ATC strata and ultimately improve the quality of minor road traffic estimates.
3. Further stratification of annual and benchmark methodology
Reviewing stratification methods has proven essential in improving the robustness of historic minor road traffic estimates as part of this historic review. The review into continuing minor road traffic estimates will investigate alternative levels of disaggregation for the measurement of the change in minor road traffic between two consecutive years. This will include whether splitting the region of London into Inner and Outer London is appropriate for the annual count point sample. It will also consider if stratification based on the flow groups calculated from the GPS dataset can be applied to the annual minor road methodology.
4. Within-year annual minor road traffic estimates
The review will consider if it is feasible and robust to produce the annual minor road estimates by using a method similar to the once-a-decade benchmark process. This approach would produce an annual estimate of traffic on minor roads by taking the traffic counts for a given year and weighting them to produce an estimate of vehicle miles travelled on all minor roads in Great Britain for the given year. This differs from the current approach, which uses the change in flows between two consecutive years to produce an estimate of the traffic.
5. Further implementation of the GPS dataset
Use of the GPS dataset has improved the robustness of minor road benchmarking methodology. In addition to exploring its use in the annual stratification, analysis of this data source could also be used to support validation of minor road traffic data. Further investigation will also consider if the GPS dataset can be used to identify relative traffic levels across minor roads, across years, or within local authorities.
Annexes
Annex A: Details and limitations of the GPS dataset
The ‘GPS dataset’ detailed in the ‘areas of exploration’ section of this document is used by the department to monitor road congestion and journey time reliability as part of the production of road congestion and travel time statistics. The ‘Travel time statistics background quality report’ available on the road congestion and travel time statistics information page gives detail regarding the source and regular use of this data.
There are limitations to the data that was available for use as part of this minor roads review which have been considered when revising historic methodology.
The dataset does not cover road links in Scotland, therefore the methodological developments for which this dataset was necessary could not be applied to minor road traffic estimates in Scotland.
The department does not hold the referenced GPS data for 2009. Instead, GPS data from 2011 was used as the closest to the 2009 benchmark exercise. Note that this dataset was used as an indication of how busy minor road links were, relative to each other. Specific observations from this dataset were not used to calculate estimates of minor road traffic.
The department holds the referenced GPS data for minor road links in Wales for 2016 only. Analysis comparing annual average daily flow (AADF) in 2009 and 2019 with GPS observations for Wales in 2016 showed a sufficient correlation to use 2016 GPS observations in the stratification of the 2009 and 2019 benchmark exercises.
The dataset gives excellent coverage of the Strategic Road Network and Local ‘A’ roads, guaranteeing over 95% coverage every day. Minor roads do not have the same level of coverage. Coverage for ‘B’ roads and ‘cul-de-sacs’ was not as high as for other minor roads. Therefore, it was not deemed appropriate to use this dataset for stratification of ‘B’ roads and ‘cul-de-sacs’.
Table 4: Proportion of minor road links covered by the GPS dataset by year and road category
| Year | Road category | Proportion of road links covered |
|---|---|---|
| 2009 | B roads | 8% |
| 2009 | Cul-de-sacs | 25% |
| 2009 | All other minor roads | 79% |
| 2019 | B roads | 84% |
| 2019 | Cul-de-sacs | 29% |
| 2019 | All other minor roads | 88% |
Annex B: Annual methodology for calculating traffic estimates for Great Britain
Road traffic estimates are currently published for Great Britain on an annual and quarterly basis, as well as an annual publication of street-level traffic data via the traffic counts website.
The scale of the road network in Great Britain means it is not possible to count traffic on every stretch of road every year. Instead, a representative sample of road sites are counted each year.
Quarterly estimates are calculated on a panel sample approach, with traffic data collected continuously from a fixed national network of around 300 Automatic Traffic Counters (ATCs) which count flows and classify by vehicle type.
Annual estimates are currently based on around 8,000 manual counts, where trained enumerators count traffic by vehicle type over a 12 hour period. This data is combined with the ATC data and road length statistics to produce the number of vehicle miles travelled each year by vehicle type, road category, and region.
For major roads (motorways and ‘A’ roads) a rolling Census approach is taken to manual counts, which enables road-level traffic estimates to be produced for these road types.
For minor roads a panel sample approach is taken, whereby the same roads across Great Britain are counted each year (over 4,000 locations). This enables robust national level minor road traffic estimates to be produced.
For the minor road traffic estimates, the sample of minor road locations remains fixed for around 10 years. Change estimates from a fixed sample may drift over time and the sample may vary and become less representative of the changing minor road network. To account for any error incurred in the fixed sample, the sample is revised through the minor roads benchmarking exercise every decade. The revised sample for each benchmark consists of around 10,000 minor road locations, with around 4,000 of these locations forming the annual minor road sample for the following decade.
More detailed explanations of the current methods used to produce traffic estimates, from the above data sources, can be found on the road traffic statistics information page.
Annex C: Minor roads benchmark methodology
This annex replaces previous technical papers on benchmark methodology.
C1. Benchmark sample
Due to the large number of minor roads in Great Britain it is not possible to count the traffic flows on all of them. Therefore, a representative sample of points (count points, CPs) on minor roads were selected for each of the minor road traffic benchmarking exercise.
The sample CPs were selected from the sample population, which for the 2019 exercise was all carriageway road links of minor roads in Great Britain as at September 2017. These are contained in the Ordnance Survey Mastermap Highways Network product, where each road link is given a unique Topographic Identifer (TOID). In the 2009 benchmark exercise, the sample was drawn from the Ordnance Survey MasterMap Integrated Transport Network for November 2007.
The samples were stratified by local authority, road classification and urban or rural classification.
C1.1 The sample design
The 2009 benchmark exercise sample was drawn with probability proportional to the length of the minor road segment (TOID length). A total of 10,483 CPs were selected. The 2009 benchmark only covered England and Wales.
For the 2019 benchmarking exercise, the total number of CPs was set at 10,000; the maximum number covered by the budget set for the project. Traffic counts at CPs can be cancelled, rescheduled or deemed unusable, so to account for this the sample was over allocated by 500 sites (n=10,500). Further detail about the 2019 sample design is set out below.
C1.1.1 The Neyman Allocation
To optimise the precision of the allocation for the given sample size, the number of CPs in each strata was determined using the Neyman Allocation formula:
The strata used were local authority, road classification and the urban or rural location. Following this initial allocation, a regional adjustment was applied to correct for low allocations and minimum/maximum allocations within the strata, as follows.
C1.1.2 Regional adjustments and minimum values
For the 2019 exercise, Wales and Scotland both had a maximum allocation; 500 counts in Wales and 730 in Scotland, the remainder were allocated to England (9,270). The regional adjustment was applied to firstly ensure that there was a minimum number of 50 CPs in each strata and secondly to ensure that this adjustment for minimum numbers within each strata would not result in exceeding the total allocation to each country.
C1.1.3 Cul-de-sac allocation
Once the Neyman allocation and regional adjustments were applied, a total of 70 counts for the 2019 exercise were initially allocated to cul-de-sacs. Cul-de-sacs tend to be very low flow roads and the behaviour of traffic flows on cul-de-sacs is very different to other roads. Therefore, it was important to include a sample of these roads in the benchmarking exercise.
C1.2 Sample selection
To select the count point locations, TOIDs were systematically selected from the sample population database of TOIDs within each strata. TOIDs were ordered by length, to help control the length distribution of the selected TOIDs, and the cumulative length was calculated starting with the shortest TOID. The total length was used, alongside the number of TOIDs needed in the sample strata, to give a selection interval (in). The first TOID was selected at a random start point, the next TOID had a cumulative length equal to that of the start point plus in. This continued until the count allocation was filled and created a probability of selection proportional to TOID length.
C1.2.1 Selection interval calculation
The mid-point of each selected TOID was given as the CP. Theoretically there are an infinite number of potential CPs on a TOID, but the mid-point is chosen to give an average level of traffic across all possible CPs. Where possible all counts were carried out at the mid-point, this was only moved due to lack of a suitable observation point or safety concerns.
The proximity of the CPs to other sampled CPs was assessed. CPs were replaced when they were within 100 metres of another sampled CP in the same strata. CPs were also replaced before and during the count season for various reasons, including: safety concerns, very short road links and incorrectly classified roads. In these cases, the next TOID in the same stratum was chosen as the replacement, aiming to select a replacement with similar characteristics.
C1.3 The final sample
C1.3.1 2009 final sample
For the 2009 exercise, which only covered England and Wales, the total number of benchmark CPs scheduled to be counted during either the 2008 or the 2009 count season were 10,483. Of these, 756 were excluded during the 2009 benchmark exercise. As part of the minor road traffic estimate review, a further 46 CPs were excluded.
Table 5: The number of 2009 benchmark minor road counts carried out in Great Britain by region
| Region / Country | Number achieved in sample | Percentage of total in sample |
|---|---|---|
| North East | 516 | 5.3% |
| North West | 1,115 | 11.5% |
| Yorkshire and the Humber | 886 | 9.2% |
| East Midlands | 795 | 8.2% |
| West Midlands | 1,141 | 11.8% |
| East of England | 1,280 | 13.2% |
| London | 741 | 7.7% |
| South East | 1,799 | 18.6% |
| South West | 1,023 | 10.6% |
| Wales | 385 | 4.0% |
| England and Wales | 9,681 | 100.0% |
C1.3.2 2019 final sample
For the 2019 exercise, the total number of benchmark CPs scheduled to be counted during either the 2018 or the 2019 count seasons were 10,767. Of these, 567 CPs were not included in the final sample, 5% of the total, these CPs were excluded for reasons such as enumeration errors and safety concerns of manual enumeration. This level was approximately what had been expected, so the final achieved sample of 10,200 CPs was sufficient. Overall 1,229 of the CPs in the final sample were replacements of original CPs, 12% of the final sample. Of the final 10,200 counts 123 were carried out on cul-de-sacs, this was higher than the initial 70 counts allocated to this type of road.
As shown in table 6, the number of counts achieved varied by region, but the proportion of counts in each region was similar to the regional proportion of total minor road length. The South West had the highest proportion of counts at 13.6%, this is similar to the proportion of minor road length in the South West (13%). The North East had the smallest proportion of counts at 4.9%, and this was also similar to the proportion of minor road length that it accounts for (4.2%). The initial allocation aimed for 500 counts in Wales and 730 in Scotland, 511 count in Wales and 725 in Scotland were achieved. These numbers were sufficient for the benchmark traffic estimates to be calculated.
Table 6: The number of 2019 benchmark minor road counts carried out in Great Britain by region
| Region / Country | Number achieved in sample | Percentage of total in sample |
|---|---|---|
| North East | 502 | 4.9% |
| North West | 1,155 | 11.3% |
| Yorkshire and the Humber | 976 | 9.6% |
| East Midlands | 882 | 8.6% |
| West Midlands | 906 | 8.9% |
| East of England | 1,276 | 12.5% |
| London | 566 | 5.5% |
| South East | 1,312 | 12.9% |
| South West | 1,389 | 13.6% |
| England | 8,964 | 87.9% |
| Scotland | 725 | 7.1% |
| Wales | 511 | 5.0% |
| Great Britain | 10,200 | 100.0% |
C2. Estimation Method
Traffic counts were conducted at minor road locations in each of the benchmark exercises. The following calculation methods were applied in both the 2009 and the 2019 benchmark exercises.
C2.1 Converting count data to daily flows
At each CP in the sample, a 12 hour traffic count was undertaken. These counts used the standard manual count methodology, as used for all manual counts undertaken by DfT. The results of these 12 hour count surveys were converted into an annual average day’s 24 hour flow, the AADF. This conversion to an AADF used the standard process of applying expansion factors derived from automatic traffic counter (ATC) data. The expansion factor applied was dependant on the expansion factor strata and the date that the count was carried out on, as it adjusts for traffic variation across the year. The annual methodology document on the road traffic statistics information page gives further information about expansion factors.
C2.1.1 AADF calculation
As traffic counts took place over 2008 and 2009 for the 2009 benchmark; 2018 and 2019 for the 2019 benchmark, the relevant expansion factors for the given year counted were applied. Following the conversion to an AADF, growth factors were then applied to the 2018 earlier year’s AADFs. This process resulted in a set of AADFs all at 2009 or 2019 levels, making the resulting traffic estimate more comparable to the 2009 or 2019 annual traffic estimate. The annual methodology document on the road traffic statistics information page gives further information about growth factors.
C2.1.2 Growth factor application
C2.2 Traffic estimation by vehicle type
DfT produce traffic estimates by vehicle type, but when assessing the variation in the AADFs by vehicle type within the benchmark strata, it was deemed inappropriate for traffic estimates using the benchmarking sample to be carried out by separate vehicle type. Specifically, the variance of bus and HGV counts were much higher than that for the ‘all motor vehicle’ total. As a result, the benchmark traffic estimate and the benchmark adjustment are on an ‘all motor vehicle’ basis, i.e. the same adjustment will be applied independent of vehicle type.
C2.3 Design weighted initial traffic estimate
The initial traffic estimate for a given strata was calculated by aggregating for all CPs. For example, using the 2019 benchmark: the 2019 AADF multiplied by the 2019 length of the TOID that the CP is on, the number of days in the year and the design weight.
The design weight was used to reverse the probability of selection for a given CP. For example using the 2019 benchmark: it was calculated by dividing the total strata length at the point of selection (2017), by the original 2017 length of the TOID that the CP is on, multiplied by the number of CPs originally selected in 2017 in the strata.
C2.4 Traffic calibration to full minor road network
Over time the total length of roads can vary, with new roads built or changes to road classifications. For example, for the benchmark traffic estimate in 2019, the original sample was based on the road network as at September 2017, but the traffic estimation calculation was as at 2019 using the achieved sample. Therefore, a calibration factor was calculated to account for this change in both benchmark exercises, using the methods set out below.
C2.4.1 Calculation of road length calibration factors
Each CP’s design weight was multiplied by the TOID length that the CP is on, and then when subsequently summed within the post-strata, this gave the road length estimated on. If road length had not changed between selection and estimation then this would equal the final road length within the strata. In many cases the TOID length will vary slightly, plus a few CPs were also lost, reducing the road length estimated on. Therefore, dividing this number by the final road length within the strata gave a calibration factor.
During the calibration steps, in addition to region, road classification and urban or rural classification, the post-strata took into account whether a road was a cul-de-sac. Below is how the calibration factors were calculated. This calculation was used to get calibration factors for both cul-de-sac and non-cul-de-sacs.
C2.4.2 Regional cul-de-sac adjustment factors
The number of cul-de-sacs included in the sample was disproportionate to the total road length that they accounted for. As a result, to offset the impact of this under sampling, the traffic on cul-de-sacs within each post-strata was based on the national estimate with an additional regional factor applied. The regional factor was calculated by comparing regional traffic levels per meter road length for non-cul-de-sacs to the national levels.
C2.5 Final traffic estimate
To create the final traffic estimate, the initial traffic estimates the post-strata level were multiplied by the relevant calibration factors. Traffic on cul-de-sacs was halved to reflect the different pattern of traffic on dead-end roads, where vehicles are less likely to drive the full length of the road.
C2.6 Benchmark adjustment
To attain the final benchmarking adjustment, the benchmark traffic estimate was compared to the 2009 and 2019 annual estimate, which was created using the standard annual methods.
C2.6.1 Adjustment factor calculation
C2.7 Providing a consistent back series
The application of the benchmark adjustment is intended to reduce the impact of accumulated error from having a fixed sample in the standard annual methodology. To give a consistent back series with viable year on year comparisons, the benchmark adjustment was applied to the published back series using annual factors. These annual adjustment factors are a geometric series, so the benchmark point adjustment is raised to the power of the factor described below to get each annual adjustment. The annual adjustments for Great Britain can be seen in tables 7 and 8.
C2.7.1 Annual adjustment factor calculation
Tables 7 and 8 show the minor road traffic estimates for Great Britain and the benchmark adjustments as they were calculated for the revised figures following the historic minor road traffic estimates review.
Table 7: Revised benchmark adjustment factors, 2000 to 2009
| Year | Revised 2009 fixed annual sample estimate (bvm) [A] | Revised 2009 benchmark traffic estimate (bvm) [B] | Revised benchmark adjustment [B]/[A] |
|---|---|---|---|
| 2000 | 103.8 | 103.1 | 0.99 |
| 2001 | 104.7 | 103.4 | 0.99 |
| 2002 | 108.9 | 106.8 | 0.98 |
| 2003 | 109.6 | 106.9 | 0.98 |
| 2004 | 109.6 | 106.4 | 0.97 |
| 2005 | 110.6 | 106.9 | 0.97 |
| 2006 | 112.3 | 108.1 | 0.96 |
| 2007 | 115.7 | 111.1 | 0.96 |
| 2008 | 114.7 | 109.8 | 0.96 |
| 2009 | 112.4 | 107.3 | 0.95 |
Table 8: Revised benchmark adjustment factors, 2010 to 2019
| Year | Revised 2019 fixed annual sample estimate (bvm) [A] | Revised 2019 benchmark traffic estimate (bvm) [B] | Revised benchmark adjustment [B]/[A] |
|---|---|---|---|
| 2010 | 105.1 | 105.8 | 1.01 |
| 2011 | 104.2 | 105.5 | 1.01 |
| 2012 | 104.5 | 106.4 | 1.02 |
| 2013 | 104.1 | 106.6 | 1.02 |
| 2014 | 108.0 | 111.2 | 1.03 |
| 2015 | 108.9 | 112.9 | 1.04 |
| 2016 | 110.7 | 115.4 | 1.04 |
| 2017 | 112.1 | 117.4 | 1.05 |
| 2018 | 110.6 | 116.6 | 1.05 |
| 2019 | 111.2 | 117.9 | 1.06 |
Instructions for printing and saving
Depending on which browser you use and the type of device you use (such as a mobile or laptop) these instructions may vary.
You will find your print and save options in your browser’s menu. You may also have other options available on your device. Tablets and mobile device instructions will be specific to the make and model of the device.
How to search
Select Ctrl and F on a Windows laptop or Command and F on a Mac
This will open a search box in the top right-hand corner of the page. Type the word you are looking for in the search bar and press enter.
Your browser will highlight the word, usually in yellow, wherever it appears on the page. Press enter to move to the next place it appears.
Tablets and mobile devices normally have the option to “find in text” and “print or save” in their sharing or quick options menu of their browser, but this will vary by device model.
