Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Statistics is suffering a crisis of confidence. The UK has had a run of flawed economic data; earlier this month chief statistician Ian Diamond stepped down, citing health issues. Statistical reliability and interpretation has also been in the docks in court. The Royal Statistical Society waded in on the case of Lucy Letby, a former neonatal nurse convicted of murdering seven babies, noting that correlation does not prove causation.
The need for reliability is axiomatic; economic stats underpin everything from interest rates to government policy to corporate investment. That is why, for instance, the World Bank spent $700mn over the past four years helping poorer countries in their statistical endeavours.
Yet over a decade after sowing the seeds of the Greek debt crisis, and crucifying inflation-linked bond investors in Argentina, where the government heavily understated inflation, inaccurate or manipulated data is proliferating.
Funding can be an issue. The US is in retrenchment mode, cutting surveys and reducing sampling sizes. An inquiry into issues at the Bureau of Labor Statistics found tech modernisation was held back by underfunding. More of that is coming down the line.
In Britain, the Office for National Statistics missed nearly a million workers from 2019 to 2024, reckons the Resolution Foundation, a think-tank. Quality issues, stemming in part from respondents’ growing apathy over surveys, compounds the ONS’ issues. The UK’s labour market data set lost its accreditation as an official statistic at its low point in 2023 after the response rate fell to 17 per cent.
Launching yet another review, as the UK is doing, seems pointless. Many of the findings from the Bean Report, published nearly a decade ago, have not been acted upon. The much-delayed Transformed Labour Force Survey is not due until the end of next year.
What to do? For surveys, bigger samples is only part of the fix. Eliminating biases is also key. Jobless people have probably more time to fill in labour surveys than workers, for example.
Answers may come from Silicon Valley. After all, nobody is better at harnessing digital sources of data, and in vast amounts. Respondents may be reluctant to engage in surveys, but information on their likes, hates, habits and transactions is given away freely to companies such as Alphabet’s Google and Meta’s Facebook.
Governments, including the UK, are starting to think about web scraping and scanning transactional data such as train tickets, or accessing the type of data held by trade publications — second-hand car markets, holiday destinations and private rentals.
Charles Bean, author of the 2016 eponymous report, reckoned the ONS’ access to raw data put it in pole position to develop the necessary alternative indicators. That might be more true, though, of the private sector. Joining the dots between the two is by no means simple, but the idea creates hope that official numbers might, one day, add up.
