top of page
User Guides

How to navigate the dashboard & understand the analysis

Our pdf user guide below provides information on how to use the dashboard and understand the analysis. 


  • ​Data Advisor: Anna Goodman - London School of Hygiene & Tropical Medicine & data lead for Department for Transport recommended Propensity to Cycle tool.

  • Guidance & insights: Alice Roberts, Head of Campaigns CRPE, London Chair, Healthy Streets Scorecards


For ease of reference, we refer to the London primary pupil school travel dashboard as "our model" in the methodology. 


1.0 School characteristics & pupil numbers.

The source data is the Department for Education (DfE) "National Statistics 2021/2022 schools, pupils and their characteristics" [school census]  release here.  We modified this data in the following ways to provide the relevant input into our model. 

  • Out of Scope:  All SEND schools and pupil referral units because they have unique travel requirements. State funded nurseries because their catchment information is unknown. 

  • Modifications to data: May 2023 changes to ward & borough local authority boundaries were applied to schools. 'All through' schools were split out into primary & secondary schools. In the school census "all-through" schools 3-18 are classified as secondary schools. These have been split out into separate primary & secondary schools & information on the primary & secondary age numbers has been taken from the Good Schools Guide where not provided in the school census. These splits mostly related to independent schools but did include some state schools too. 


Note - some schools that have two sites may be listed as one site where they are treated as such in the school census. 

2.0 Pupils travelling within distance bands 1 mile, 1-2 miles and  2+ miles from their school - modelled data


2.1 Pupil residence distributions 

We think pupil travel distance band information is really valuable information for authorities to provide, but as of December 2023 we had not been able to find any authority that had this information calculated and available for release having tried both the Greater London Authority and the Department for Education (DfE).  We have therefore modelled this information using existing catchment information available in the following way:


Catchment area inputs - we used publicly available pupil catchment datapoints on the distribution of pupils in reception. This originated either from the DfE 2021/2022 school census data or in a minority of cases from local authorities 'last distance offered' releases which we then modelled to a reception pupil distribution (last distance offered does not usually account for sibling priority etc). These catchment inputs were an aggregation of reception pupils for each school, providing information on the distances that 50%, 70% and 90% of reception pupils live from their school (for example) or the furthest distance that a reception year place was offered. They were also "as the crow flies" distances which measure how far a pupil lives from school by  straight line distance on the map from the school to pupil residence - the technique used for most pupil admissions. 

2.2 Converting "as the crow flies" distances to on-the-road distances. 

The distance that pupils travel on their trip to school is longer than the "as the crow flies" distance because they need to navigate the street and neighbourhood lay-out, they can't walk in a straight line across a map. In order to convert these "as the crow flies" inputs to "on-the-road" travel distances we used a conversion calculated using data available from the Propensity to Cycle Tool. For example, a crow-flies journey of under 1 mile is scaled up by 1.52 to convert into the average on-the-road distance a pupil travels, so a child who lives a 0.5 mile distance from school by straight line distance actually on average travels 0.76 miles to school on-the-road. We have included more detail about these calibration factors used below. The Propensity to Cycle data source includes all origin-destination at the level of Lower Super Output Areas (LSOAs). LSOAs are administrative Census areas containing around 1500 individuals each. In the Propensity to Cycle Tool, these origin-destination commute trips are modelled as starting and ending at the population-weighted centroids of each LSOA pair, the Propensity to Cycle Tool calculated a) the crow-flies distance between the centroids and b) the on-the-road distance between the centroids. The latter was defined in terms of the ‘fastest’ route returned by the cycle journey planner CycleStreets. Note that this fastest cycling route is generally likely to be a fair approximation for the on-the-road driving distance, although the cycling distance may sometimes be shorter (e.g. if the cycle route goes through a traffic-free park). Note also that the crow-flies distances are rounded to the nearest km and the on-the-road distances rounded to the nearest 100m in this dataset. For each of the following crow-flies distance bands, our advisor Anna Goodman – data lead for the Propensity to Cycle Tool - identified the average scaling up factors to convert to on-the-road-distance. Crowflies distance in Propensity to Cycle Tool data (miles) - Average scaling up factor to convert crow-flies distance to on-the-road-distance: Less than 1 - 1.52 1 to under 2 - 1.36 2 to under 3 - 1.31 3 to under 4 - 1.28 4 to under 5 - 1.26 5 to under 6 - 1.22

2.3 Converting aggregated pupil travel distances into travel bands. 

In order to convert our aggregated known input on pupil travel distance bands for each school, for example "50% of pupils travel distance X to the school and 70% of pupils travel Y distance to the school",  we used linear modelling to convert these known aggregated data points into the numbers of pupils travelling under 1 mile, between 1-1.9 miles and 2 miles or more. 


The pupil travel distance bands are therefore intended to be a reasonable estimate for the purposes of better understanding pupil travel modes.  These distances are estimates, though for simplicity we have not labelled them as such in our dashboard analysis as we feel they will be fairly closely aligned to reality.


Caveats; we have had to use aggregated data sets e.g. 60% of pupils live within distance X from school, and applied on-the-road scale factors to that aggregate, which will differ from applying those factors to every individual pupil distance; the on-the-road scale factors used in our model is by cycle-travel on-the-road, however walking distances may sometimes be shorter e.g. walking through an alley between roads rather than cycling on-the-roads around;  we have taken aggregated data points & used linear modelling to convert those into travel bands, which again, will vary from exact distances.  We summarise an overall interpretation of the caveats in our model in a comparison to National Travel Survey data points below - Section 4.0. 

2.4 Independent school pupil distributions  - estimated data

Independent schools don't disclose their pupil distributions publicly. In order to include them, we used the ratio of pupil distribution between a secondary state grammar school [admission process is academic selection] and a secondary state catchment school [admission process is usually closest straight line distance of pupil residence from school], and applied that to the average pupil distribution of a London catchment primary school to estimate the pupil distribution for London primary independent schools. In our model this resulted in 29% of pupils travelling under 1 mile, 29% travelling between 1-1.9 miles and 43% travelling 2 miles or more to their independent school. This compares to a  London state primary catchment where our model shows 79% of pupils travel under 1 mile, 16% travel 1-1.9 miles and 6% travel 2 miles or more to school.  Based  on our knowledge of Dulwich independent schools we believe this estimate has face validity. It is, however, an estimate and will vary by school.

Caveats; this is an estimate that has been used across all independent schools. It will not take into account individual circumstances of schools and areas. For example there are a high number of Jewish independent schools in Hackney wards which are likely to have far smaller pupil distributions because they serve the Jewish communities in their localities, versus academically selective independent schools that recruit from 3+ mile radius.

3.0 London primary pupil driving rates  

Data Source – DfT National Travel Survey [NTS] statistics. We have used London mode of travel data in the underlying NTS statistics. The years used were an average across 2017, 2018, 2019 and 2022. NTS collects travel diaries across a week for all members of participating households. These statistics are generated based on a total sample of 4,593 trips made for the purpose of 'education', by 435 children aged

5-10 living in London.

This analysis of these underlying data tables for London showed the following pupil driving rates across distance bands which applied to the numbers of pupils in distance bands within our model. 

Distance bands & % of pupil trips driven

Under 1 mile              7%

1 to under 2 miles   65%

2 to under 5 miles   67%

5  miles +                 58%

4.0 Comparatives to NTS data. 

Across the NTS years analysed (2017, 2018, 2019, 2022),  which covered 4,593 trips made by 435 children, their analysis showed that 65% of pupils travelled under 1 mile to school and 35% over 1 mile. The blended pupil driving rate across these trips was 28%.  In our model,  we estimate that a slightly higher number of  70% of pupils are travelling under 1 mile to school and 29% over 1 mile. Our blended modelled driving rate across all distance bands is therefore lower, at 25%.  The reasons for the differences between these two estimated are explained further below in Section 5.0, but we think that they are sufficiently close for users to have  confidence in using the school, ward & borough level data in our model to further their campaigns for sustainable school travel. 

5.0 Summary of caveats - reasons for variations between our modelled results & NTS data. 


5.1 Differences in profile of pupil trip lengths

As explained above the NTS has used a sample of 435 children for London,  and within that sample size, 65% of their school trips were under 1 mile. We have used catchment information inputs for almost every individual school in London & estimates for independent schools (see Section 2.0 for how these are developed) as a basis for our output that 70% of pupils travel under 1 mile.  Whilst the NTS survey is subject to the limitations of sampling approaches, our model is subject to the limitations set out in Section 2.0, namely, using aggregated catchment points to represent pupil distribution & using scaling factors to derive on the-road-travel-distances, rather than those actually reported by parents in the NTS survey.  

5.2 Differences in pupil driving rates

The NTS data has a blended driving rate of 28% of primary pupils and our model calculates a modelled driving rate of 25%. The main reason for this variation is  due to the differences in the profile of pupil trip lengths, as explained in 5.1.  Since our model shows a higher proportion of pupils travelling shorter distances, it follows that using the NTS average driving rates within distance bands, so our blended pupil driving rate will also be lower. 

However, NTS sample data does not of course show granular, school level, local insights. We think that our model is sufficiently close to NTS data for users to have confidence in using the localised school, ward and borough level data included.

5.3 Caveats at local levels

Our model is an estimate using pupil catchment information and London average driving rates. It therefore won't account for key factors that will cause variation from average driving rates such as car ownership levels, safe availability of alternative walking & cycling routes, public transport accessibility, congestion zone, school, or community culture and so on. 

6.0 Conclusion

We want to conclude by flagging that it is in theory absolutely  possible for the DfT and the DfE to track actual, individual pupil travel modes by school, and indeed by pupil distance travelled. However, the last year that school level travel mode information was collected nationally was 2011 when it was included in the school census. This is despite significant recent national efforts to improve pupil active travel to school. School run driving rates have remained almost unchanged for the last 20 years and we urgently need more publicly available, granular data on the school run to affect them.  We hope our model highlights how useful publicly available school, ward and borough level information on pupil travel modes can be & propels forward more systemic initiatives to collect it. As the saying goes "you can't manage what you can't measure".

bottom of page