We recently published a blog post on the UK’s fastest growing city economies. This generated some sceptical replies. That scepticism was justified.
At the time the analysis was done, The Data City used older OECD city definitions, and some of the repeated analysis in this blog post will generate slightly different numbers as a result. These changes make no important difference to this analysis except for the absence of Bradford which has been absorbed into Leeds in the latest OECD city definitions.
A repeat of the analysis that formed the basis of our blog post returns a very similar result. Liverpool and Sheffield are reported to be the fastest-growing large cities in the UK. This is almost certainly incorrect.
We know it is incorrect because we can check in the OECD’s data on city growth.
The Data City’s current city definitions are now perfectly aligned with the OECD functional urban area definitions. The OECD publish economic data for these cities including total GDP, the growth of which we would expect to be highly correlated with company growth.
Using this data we can rank the UK’s cities by GDP growth. We choose 2014 as the start year as this is the oldest data in The Data City platform. We choose 2019 as the end year as this avoids the variable and poorly-measured effect of the Covid recession.
Restricting ourselves to the largest 25 cities in the UK, we get the following growth rates and rankings.
These results are reasonable. The fall in global oil and gas prices in the period have led to a recession in Aberdeen. Milton Keynes, a long-standing high performer in the UK economy has continued its strong performance.
Sheffield and Liverpool have growth rates below average. Including growth to 2020 does not change the story. The fortunes and economic trajectories of large cities rarely change overnight and past performance is a reasonable guide to future performance. This data should have caused us to question our data, but there were other signs that we missed.
The second graph in our original blog post looks at the growth in the projected turnover of all companies in Sheffield. Even without checking our data against the OECD’s this should have made us question our findings. It is very unlikely, over a period where the UK economy barely grew, that companies in Sheffield genuinely saw a doubling of total turnover.
What is happening?
To try and understand what is happening with this data, we use The Data City platform. In EXPLORE we set the location filter to Sheffield and sort the companies from High Growth to Low Growth.
Immediately we see a problem. VOLTMETRIC SPECIAL PROJECTS LTD (Company Number 10551648) has an enormous growth rate and nearly 200,000 estimated employees. Given its SIC code and our turnover estimation system this gives it an estimated turnover of around £11bn. Yet I’ve never heard of it.
Looking at the company’s financial data, we see a rare but known problem. The company accounts which we hold have a clearly mis-entered employee count, a near-duplicate entry of the Net Worth figure in the row above.
With this company in the data, Sheffield’s economy in The Data City platform shows a growth rate of +11%, as in the summary table at the start of this blog post.
Once this company is removed, the estimated growth rate falls to +5.7%.
This correction alone means that Sheffield falls to a position in the bottom third of the table, consistent with the ranking of Sheffield’s growth by the OECD.
In Liverpool we find the same thing. The fastest growing company, MAGNETIC MEASUREMENTS LTD. (Company number 02978036) has an implausibly large number of employees.
Inspection of its financial data shows the same pattern of incorrect data entry.
Excluding this company from our analysis pushes Liverpool growth rank back to what we would expect given the OECD data.
The RTIC process
During the creation of each of our RTICs we perform a thorough quality assurance check. This does not guarantee perfection, but it reliably removes most large outliers such as the two identified companies here. Neither company is a member of any RTIC on our platform.
Furthermore, when working with RTIC data, the number of companies is sufficiently small that outliers that have made it through the QA process during RTIC creation are found during analysis.
The Data City’s data works well at RTIC scale, and especially at RTIC and place scale. We can manually check for data errors and remove companies that would distort analysis.
Our methods are not suited to analysing the whole economy, or even the whole economy of a single city. We have national statistics for this purpose.
The Data City advantage
We should not have published our previous blog post and we are linking to this correction from it.
Our innovative approach to understanding and measuring the economy has huge advantages over traditional approaches in many areas. We can understand what companies do in much more detail. We can see emerging trends much earlier. We can identify clusters of excellence in the economy much more rigorously.
But our data is a supplement to and not a replacement for more established economic methods. It should always add more detail while almost never significantly deviating from the picture that traditional economic statistics provide.
We work hard to document the limitations of our data and to explain the care that must be taken in its analysis before drawing conclusions from it. We work hard with our clients to get this right. In this case we failed to apply our own advice to our own work. I am working hard to make sure this doesn’t happen again.