Britain's large cities are as densely populated as their European equivalents.
, .
About 1000 people every day directly use my population around a point tool to calculate population within circles anywhere in the world. The API is incorporated into at least a dozen more apps (the API is not authenticated so I don't know beyond unique IP addresses), creating a total of 50,000 requests per day on average and up to 250,000 requests per day at peak times.
The most common feature request I get is to remove the minimum 3km radius in the tool's UI. This blog post is about why the 3km minimum radius and its associated warnings exist and the problems that ignoring those warnings can cause.
Below is a map of Leeds, my city, with a 3km radius circle around the city centre. The pink circles are the centres of the gridded population data provided by the Global Human Settlement (GHS) project which powers my tool. Their size is proportional to the population in that grid.
When I wrote the tool I lived in a twelve storey block of flats, surrounded by other such blocks. And yet the circle representing that grid square, it is the Southernmost point with an arrow pointing to it, is small.
Today I live in a terraced house, with a front and back garden, on the largest circle on the map. I am confident that I have not moved from a low density part of Leeds to a higher density one. Nor do I believe that where I live now is truly one of the densest squares of population in England.
The GHS dataset is a marvel of modern data. It produces comparable population data for countries that barely have functional statistical offices. But this does not make it infallible, especially at small scales. Britain's statistical institutions do not provide good enough data to GHS frequently enough to produce excellent output. On top of this, the GHS team do not have sufficient supplementary data and resources to fix weak input data and make fine-grain population density estimates for the centre of fast-growing British cities with a changing urban form — tall modern flats are not a typical home type in Britain and this interacts with satellite estimation methods — accurate.
The GHS team ensures that over larger scales their methods are correct even when in the detail they are not. This is why my current grid square has such an exaggerated population. This adjustment make up for underestimated populations closer to the city centre while satisfying constraints on the population of Leeds at larger geographies.
At 3km, the population of central Leeds is safer to use. Initially I set the lower bound at 5km, but this was unfair to much of the rest of the world which provides better data to the GHS team.
The second most common feature request I get for my tool is for more complex analysis forms considering ultra-local density distributions and similar. I hope that I have shown why I have not added these features and why I react poorly to accusations of being overly simplistic in doing so. Such analysis is not statistically safe in England and Wales if based on the GHS dataset.
Ring populations and national data.
To do more complex analysis of population density at smaller scales in Great Britain requires new datasets. My older, slower, more complicated, and thus largely unused ring populations tool for GB uses population estimates that are more accurate at smaller scales. They also provide data over time. These datasets are provided by National Records of Scotland (whose data is extremely fine-grained) and the ONS for England & Wales (whose data is better than that from GHS, but not as good as Scotland's).
This lets a user place concentric rings anywhere in Great Britain and calculate the population and thus the population density within those rings.
Importantly, the tool returns the change in population and population density up to mid-2024 for every year since 2002 in England and Wales and 2001 in Scotland. It uses a much more complicated, slightly more accurate, and much slower algorithm to calculate populations.
Using this tool we can show that Manchester's central density (up to 2km from the city centre) has increased by a factor of three in the last two decades.
In 2021 I wrote a piece celebrating this densification of Manchester as an example of the English local planning system responding well to clearly expressed local preference for "brownfield first" development and urban renewal over expansion of cities onto the greenbelt.
The pattern we see in Manchester is simply the most pronounced example of something that we have seen in every large city of Great Britain, with one exception.
London's core (within 2km of Trafalgar Square) population is today, according to the ONS, the same as it was in 2006. Manchester's core has overtaken Edinburgh to become the most densely populated large city core in Britain. Anyone who has seen the rapid rise of Manchester's skyline will believe this.
Britain's cities are already dense.
Having a second source of data for calculating city densities lets us check the GHS population data and show even more clearly why using it in the centres of British cities requires great caution.
Calculating the population at a distance from the centre of Manchester using two datasets, the GHS dataset, and the ONS small area population dataset, gives more or less the same answer at 5km and beyond, but the core densities are massively underestimated.
The Global Human Settlement gridded population dataset underestimates the number of people who live within 2km of central Manchester by 40%. Extrapolating the historic trend growth rate for this area of Manchester would make this nearer 50%.
This is why I don't provide a UI to easily calculate populations for circles smaller than 3km in my population around a point tool. It is also why I warn against drawing small circles in the centres of fast-growing cities.
This problem with the GHS is important. Low density of large British cities is argued by many to be a substantial source of the lost agglomeration benefits which leave large British cities except London among the poorest in Europe.
When I show, using my population around a point tool, that Manchester has as many people living within 8km of its city centre as far richer Copenhagen, Amsterdam, or Hamburg — or that Leeds has as many people living within 8km of its city centre as far richer Dublin or Rotterdam — I am told that it's really core density that matters. I am assured that if my warnings about the errors in the GHS dataset are ignored, my own tools will prove it to me.
What if instead of ignoring warnings we fix the data for Britain's large cities? I have calculated the corrected best estimates of population density in the core (<2km) of some of Great Britain's large cities from the more accurate and fine-grained ONS and NRS data.
This changes the story significantly.
London's core shrinks to only middling in Britain. Manchester and all other English big cities see their core populations rise substantially. And with cities like Leeds having built new central flats at a record rate, and having a record number of housing units under construction, the gap between the GHS estimates of core population and reality will grow. Britain's large cities are already as densely populated as many of their European equivalents.
Comparing British cities against their European and North American equivalents using these real core density numbers shows that the gap has now been mostly made up.
Manchester's city core is as dense as Antwerp, Rotterdam, or Cologne. It is even similar to more Southern cities such as Bordeaux, Milan, and Lyon, though making such comparisons in a city where you can't enjoy a drink on an outdoor terrace or balcony for most of the year is not sensible. There is a reason why people in Northern Europe and Northern America live at similar, lower, densities than people in Barcelona when they can.
The population density of the city core of Leeds is similar to that of Chicago or Dresden.
The remaining gaps between British cities and their equivalents will be quickly made up as recently completed central core homes enter the data, as new core homes under construction are completed, and as core homes already given planning permission are built out.
Still little growth.
Increased graduate share in Britain's large cities except London did not substantially improve their economies in the way that many British economists expected.
Having just about achieved comparable population densities to Northern European and Northern American comparator cities, all with vastly stronger economies, I do not believe that further densification of Britain's large cities will deliver substantial economic overperformance. What will matter much more and what already works in equivalent cities with similar densities and much stronger economies is,
- Investing enormously in improved transport, especially public transport, to enable agglomeration benefits. Rotterdam has 5 metro lines, 9 tram lines, and a fully electrified railway. Leeds has none of that. Copenhagen has 4 autonomous metro lines and 7 fully electrified S-tog lines which are being converted to autonomous operation. Manchester has none of that.
- Public sector investment in research & development which has thus far been focused in London at the expense of the rest of the country.
- Substantial fiscal devolution so that British cities can raise taxes, borrow money, and invest in their own success when the national government won't.