I’m finding this new map of life expectancy across London’s Tube lines rather interesting:
It works from a background of life expectancy calculated for each electoral ward, and then interpolates to give a life expectancy estimate for a person born within 200 metres of each Tube station. It’s a very handy thing to do, because while the electoral ward (populations of 5,200 to 23,450, with a median of 12,500) is the finest-grained level at which London makes these sorts of statistics available, it’s not a particularly intuitive unit to most Londoners. The Tube map, on the other hand, is intuitive to most Londoners, and by combining the familiar line colours and station names with a geographically-correct layout, they’ve made something that almost anyone familiar with London can read easily. This lets us readily see two important things: there is troubling inequality across London, in health outcomes as well as in income, but it’s not distributed along the simplistic East-West, North-South or Inner-Outer axes that people tend to talk about. It’s much more of a patchwork, with some rather small islands of success and failure jumbled up with each other, in a way that’s quite hard to describe concisely.
That said, it does run into some issues with the underlying data. I started digging into this because some of the West End numbers seemed really odd: how could the life expectancy around Oxford Circus station possibly be as high as 96 years, and how could the half mile between there and Tottenham Court Road cause a drop-off of 11 years?
Fortunately, the Greater London Authority has done a great job of making this kind of data publicly available. There’s a detailed list of life expectancy estimates for every ward in successive years, complete with confidence intervals, downloadable from data.london.gov.uk. I downloaded that and started playing around, and I’m now convinced that the map shows up some irregularities in the data. For the 2005-2009 reporting period (the latest available, so presumably what the mapmakers used), the West End ward that contains Oxford Circus has an average male life expectancy of 93.21 and female of 99.55. I think the 96 on the map must have come from averaging these two. But when I checked the previous period (which is 04-08, so heavily overlapping), I see a male life expectancy of 81.31 and female of 84.29. If these figures had been used instead, Oxford Circus would have been listed as 82 or 83, much more in line with other affluent areas of town.
I’m not exactly sure why this should be, but my best guess is that it’s based on too small a sample. The population of that ward isn’t terribly small, at 9,850, but life expectancy statistics are calculated directly from death rates, and something as simple as moving a hospital, hospice or retirement home can have a pretty big impact on death rates at this local a scale. Or it could even be as simple as the raw number of deaths in a small, affluent population being so low in the first place that a change that big is all noise.
I don’t mean this as some terrible criticism of the map, which is still a useful tool. But it does serve as a reminder of a few things:
- If stats don’t pass the “smell test” they are always worth investigating further.
- It’s extremely difficult to get robust data about small areas and groups of people.
- Although most people will only ever look at things like the map, not the raw data, it’s important to publish raw data so it can be checked.
- Every statistical portrait like this represents a snapshot in time, so even if there’s nothing wrong with the numbers themselves there’s always a risk that they accurately depict an unrepresentative moment.