Where shared commuting makes sense in the UK

Why I built it

I work on a commuter carpooling product in Lagos, where the question every day is "who is driving the same way, and could they share?" I wanted to ask the same question of a country with excellent open data. The 2021 Census publishes, for every local authority, how people travel to work and how far they go. That is enough to find the places where lots of people drive alone over distances that a shared car could cover.

The report has four sections, and this page follows the same order: the landscape (how people commute), the divide (London against everywhere else), the distance layer (how far they go) and the opportunity (where sharing could work).

The landscape

The report covers 19 million commuters across 331 lower-tier local authorities (the district and borough level of local government). They fall into five types of area, which I use as the main slicer throughout.

Nationally, 65% of commuters drive to work alone. The landscape page lets you pick a geography type and see the mode mix area by area, as a 100% stacked bar for every authority in that group.

One limitation sits over the whole report. Census day was 21 March 2021, during a national lockdown. About 31% of workers nationally worked from home that day, far more than in a normal year. I exclude work-from-home from every mode share, so the shares describe the people who did travel. That keeps areas comparable, but it is still a snapshot of an unusual moment, not normal commuting.

How to read the summary tiles

The landscape page has a row of tiles above the table. When London Borough is selected, the "% Drive to Work" tile shows 33.4, while the cover page says 65% drive to work alone nationally. These don't contradict each other. The tiles follow the slicer on the page, so 33.4 is the London figure for that selection, not a national one; change the selection and the tile changes with it. Only the "Total Commuters (National)" tile ignores the slicer. The same applies to the "% Work From Home (Mar 2021)" tile (43.3 for London, against about 31% nationally).

The divide

The clearest pattern in the data is the gap between London and everywhere else. In the London boroughs, about 41% of commuters use public transport. In the rest of England and Wales it is about 6%. The car takes up the difference: on the divide page, the solo-driver bar for the rest of the country is roughly twice the London one.

A scatter of all 331 areas, car-driver share against public-transport share, shows the trade-off directly. London boroughs sit high and to the left. Almost every other area sits in a tight cloud in the bottom right, with public transport under about 20% and solo driving mostly between 60% and 80%.

To make that cloud easier to talk about, I put every area into a car-dependency band based on the share of commuters who drive alone.

In 54% of all local authorities, 70–80% of commuters drive to work alone. Add the 60–70% band and nearly four in five areas are heavily car-dependent. That is the pool where a shared car is most likely to be the realistic alternative, because in most of these places there isn't a train or bus that does the same job.

The distance layer

Mode alone doesn't tell you whether sharing would work. A 3 km drive is better replaced by a bike, and a 60 km one is a hard daily commitment for a driver to share. So the third section adds TS058, the Census distance-travelled-to-work table, grouped into short, medium and long distance shares for each area.

Reading the distance-profile chart, London and the rest of the country have similar short and medium shares (each roughly 40–48%), but long-distance commuting is far more common outside London: roughly 18% of commuters against about 7%, read off the chart. People outside London don't just drive more; more of them drive further.

Two more views on this page explain London's numbers. A "commuter belt" chart ranks the home-county districts by train use and by the share travelling 30 km or more; places like Brentwood, Sevenoaks and Basildon combine high train use with long trips into the capital. And within London, plotting Underground use against car use shows a substitution effect: the inner boroughs with good Tube coverage, such as Newham and Tower Hamlets, have low car shares, while outer boroughs such as Bromley, Bexley and Kingston upon Thames have little Underground use and much more driving.

The opportunity

The last section brings the two signals together. Shared commuting has the most room where car dependency is high (lots of people already driving, few alternatives) and journeys are mid-range (long enough that walking, cycling or a short bus ride don't work, short enough to share every day). I combined the car-driver share and the medium-distance share into an opportunity score for each area, rescaled from 0 to 100 so the top area scores 100.

The top ten areas are all non-metropolitan districts, and all ten score above 90:

1. South Staffordshire (the top score, 100)
2. Bromsgrove
3. Staffordshire Moorlands
4. North East Derbyshire
5. South Derbyshire
6. North Warwickshire
7. Broadland
8. Blaby
9. Ribble Valley
10. South Norfolk

My reading of the list: most are semi-rural districts within reach of a large city (Birmingham, Derby, Sheffield, Leicester, Norwich, Stoke, Preston). People live in villages and small towns and drive into the city or its edge every day. Those are the corridors where several people make the same mid-range journey alone, and where a matched shared car could save real money.

The companion scatter, medium-distance share against car-driver share coloured by geography type, makes the same point: the non-metropolitan and Welsh areas cluster in the top right, and London sits on its own to the left.

How I built it

1. Getting the data. I downloaded TS061 and TS058 from Nomis, the ONS labour-market and Census data service, as bulk CSVs at lower-tier local authority level for England and Wales.
2. Shaping in Power Query. The Nomis files are wide, with long Census category names as column headers. I renamed the columns to short, consistent names, kept one row per local authority keyed on the ONS area code, took the work-from-home count out of the travelling total, and grouped TS058's distance bands into short, medium and long. The geography type can be read straight from the ONS code: E09 for London boroughs, E08 metropolitan districts, E07 non-metropolitan districts, E06 unitary authorities and W06 for Wales.
3. The model. A simple one: the mode table and the distance table joined one-to-one on the area code, plus the geography type and a car-dependency band column for slicing. One row per area keeps every visual honest about what a "row" is.
4. Measures. DAX measures for total commuters and each mode's share of people who travelled, built as ratios of sums so they stay correct when you filter to a group of areas, plus the opportunity score and its 0–100 rescaling.
5. Report design. A summary page that states the data and the limitation up front, then one page per question, with navigation buttons, and a published-to-web version so anyone can explore it without a Power BI licence.

The pattern for the share measures is the same throughout: sum the numerator and the denominator over whatever is in the filter context, then divide.

% Drive to Work =
DIVIDE ( SUM ( Commute[car_driver] ), SUM ( Commute[travelling] ) )

Why this matters to me

This is the same question I work on in Lagos, asked of a different country. For Conductor, a commuter carpooling platform, I mapped where Lagos commutes to find the corridors where drivers and passengers overlap, and built the model for pricing a commute so that a shared seat costs less than a solo ride. The UK report applies the same logic to open data: find high car dependency on mid-range journeys, and you have found the places where sharing has room to work.

Caveats and what I'd do differently

• A pandemic snapshot. Census day fell during lockdown. Excluding work-from-home keeps areas comparable, but the people still travelling in March 2021 were not a typical mix; key workers and people who couldn't work from home are over-represented. The 2011 Census, or the Department for Transport's National Travel Survey, would be a useful cross-check.
• Averages of averages. Some visuals use Power BI's default aggregations ("Average of", "Sum of") on area-level percentages. "London against the rest" is then an unweighted average of areas, so a small district counts as much as Birmingham. I'd replace those with explicit, commuter-weighted measures everywhere.
• Bus is inside public transport. The landscape table shows bus as its own column next to public transport, so the row adds to more than 100%. In the stacked bar it is stacked on top too. I'd show bus as a breakdown of public transport, not alongside it.
• Check the outliers. On the opportunity scatter, a small group of non-metropolitan districts shows a medium-distance share above 70%, well apart from the rest. That could be real, but it is exactly the kind of point I would trace back to the source rows before trusting it.
• Local authorities are big. A district contains both towns and villages. The opportunity score says where to look, not which streets; the next step would be the same analysis at Middle Super Output Area level (small areas of roughly 5,000 to 15,000 people), with origin–destination flows to find the actual corridors.
• "Room for sharing" is not demand. High car dependency on mid-range trips is a necessary condition, not proof people want to share. A survey or a pilot would be needed to test it.

All figures on this page come from ONS Census 2021 tables TS061 and TS058 (England and Wales), published under the Open Government Licence. Where a figure is read off a chart rather than a label, the text says "roughly" or "about".

Skills shown