The Tech Map London is out, and so is the research that underpins it. It’s an extremely impressive piece of work, and anyone remotely interested in urban tech ecosystems should take a look. Kudos to the GLA for commissioning it, and to Trampoline Systems and SQW who put the thing together. Other city-regions should try and do something similar.
Here’s some notes I made. It gets a bit geeky in places.
1/ Patterns – one widely-reported headline is that London has a whole bunch of technology hotspots, not just one. That makes sense, and chimes with other recent analysis. And as some colleagues and I explore in a forthcoming paper, even pre-2010 Silicon Roundabout was linked into a much larger system.
Another is that the tech sector is ‘shunning’ the Old St area. That’s harder to see, as there’s no time dimension in the data, but it’s clear that as that neighbourhood’s technology scene grows, and the area gets pricier, things will tend to spread out. This is what I found recently in some work for Centre for London.
2/ Definitions – The definition of ‘tech’ is important, and this NESTA piece makes clear, there’s a bunch of competing definitions in play. The project team base their work on the recent ONS science and technology categories, though they tell me they tweaked these a bit. This feels sensible, and has the advantage of allowing them to consider (say) medicine and life sciences alongside ICT.
3/ Data – the report uses high quality IDBR data for some of the analysis, but relies on Companies House data for the actual mapping, which identifies tech firms using self-reported industry codes. This isn’t great, as the authors acknowledge: a non-trivial share of firms don’t report anything, others put down non-informative codes (say, ‘other business services’), and SIC codes often don’t tell us much about products/services. Companies House data on employment and revenues is also quite gappy, and comes off of a selected subsample. Use those numbers with caution.
Anna Rosso and I have used a big data-driven approach [unlocked version] to try and get around some of these issues, though this isn’t perfect either. We’re now testing a combination of administrative and modelled info which should plug a lot more of the holes.
4/ Location – I’m still scratching my head a bit on this. Companies House data gives the address of a registered office, not the trading address. The two could be quite different, and in extremis, not even in the same city. The project team did a survey to explore these issues, finding that for most SMEs, the two addresses are the same, so developed the map on that basis. It’s obviously critical that the survey is robust for us to believe the map.
I couldn’t find that much detail in the report, but assuming the survey is sound, this is a pretty helpful finding for me and others working with company data. Meanwhile, we can get a rough sense of the correspondence by comparing the map at the top of the page with this one.
The first uses IDBR employment data from actual plant locations, the second uses Companies House registered addresses. For some reason, the first map covers the whole of science and tech, while the second only looks at digital technologies (around 18% of all science and tech jobs in 2013) and is in logs, not raw counts. The two line up *fairly* well, but really we need to see a like-for-like comparison using plants/enterprises, not jobs. Note: I’d be very happy to update this material if the team can furnish me with more detail.