Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Sherlock Holmes, the famous fictional detective, proclaimed that “it is a capital mistake to theorise before one has data”. What is true for making criminal deductions, is also true for making policy. But a problem arises when the data is not entirely trustworthy to begin with — something the Bank of England knows only too well. The Monetary Policy Committee, which held interest rates at 5.25 per cent at its meeting on Thursday, has been trying to make decisions during this rate-raising cycle with what it says is increasingly uncertain labour market data — a major determinant of inflation.
Britain’s Office for National Statistics had to publish patchy and experimental jobs data last month as low response rates for its Labour Force Survey raised the risk of bias. It has fallen from 50 per cent a decade ago to 15 per cent now. The survey is also based on outdated population estimates. The troubles with employment numbers follow significant revisions by the ONS to the UK’s gross domestic product data too. Those updates revealed that the economy was no longer likely to be a G7 outlier with the worst post-pandemic recovery.
Bad statistics lead to bad decisions. For central bankers, setting interest rates and making economic forecasts are hard enough in a period of uncertainty, let alone when the underlying data is flimsy. It matters also for the government. Chancellor Jeremy Hunt announced measures to tackle Britain’s rise in worker inactivity in the spring Budget. Issues with the LFS now raise questions over the nature and extent of the problem. Businesses also stand to lose money by making judgments on flawed statistics.
Trust in official data is essential for public discourse. Different viewpoints can otherwise be formed depending on which source is used. And unreliable statistics also undermine confidence in expertise.
For all the bad publicity, the ONS remains a respected international authority on data. The pandemic and declining survey response rates have been a challenge for statistical agencies everywhere. The ONS was in fact considered to be a world leader in tracking Covid-19. But it is precisely because of its integral role in analysis and policymaking — and the demand for more and faster data — that improvements are essential.
First, the ONS needs to review how it sets priorities and allocates resources. On Thursday, it highlighted ongoing efforts to improve its LFS data collection and methodology. It expects to transition to a Transformed Labour Force Survey early next year. Nonetheless, reasons for its sluggishness in reforming the survey, given the decade-long decline in response rates, need to be understood.
Second, it should accelerate existing efforts to improve its data. This includes exploring new ways to incentivise respondents and utilise the latest technology — it is only just shifting towards online first surveys with the TLFS. More diverse and real-time data sources can also support GDP and labour market modelling, particularly when surveys are insufficient or other data is delayed. This requires greater co-operation from government bodies to share more data, faster, to help avoid major revisions like those to GDP.
Third, the agency must attract the best talent. Data scientists are often drawn to roles in the higher-paying private sector in London, well away from ONS headquarters in south Wales. And finally, it must improve its communications; a clearer articulation of the confidence bands around its statistics would help.
The ONS is already acting on many of these. But it deserves more support from the government if it is to maintain its international standing. It has already proven itself to be efficient with a tight budget; so any squeeze on it should be reconsidered. After all, when major economic decisions are at stake, the benefits of better data can be priceless.