Yesterday the Bank of England released its quarterly Monetary Policy Report (MPR), with their latest macroeconomic forecasts, alongside their latest decision on rates.

It was a 6-3 decision to keep Bank Rate at 5.25%, with the three votes against being for an increase to 5.5%. This reflects a broad view that inflation threats remain high, but that there should be time for the already-implemented increases to transmit to the rest of the economy.

The Bank’s forecasts for GDP are not especially rosy – they weren’t in August either, but have now been slightly downgraded again, with no growth forecast in the whole of 2024, before a slightly slower recovery than anticipated in 2025.

Of course GDP forecasts can differ between official bodies – it’s a very judgement-heavy exercise – but if the OBR were to take a similar view, this would be bad news for Chancellor Jeremy Hunt in less than three weeks’ time as he presents the Autumn Statement. According to the OBR’s official timetable, the ‘pre-measures’ forecast – the last one to take into account changes other than policy decisions – was finished on 31 October, meaning that market determinants will have been taken very close to the Bank’s for the MPR, so it wouldn’t be surprising if they landed in a similar place.

The Bank sees a less tight labour market… But the signs are slightly blurry

One of the Bank’s justifications for keeping rates steady was the sign that employment and pay growth might be easing in the last few months. But the picture is not extremely clear – and the statistics certainly aren’t.

As has been extensively covered, the ONS was forced to withdraw its headline labour force statistics in October. The organisation has come under heavy criticism for it, although some of the ‘all our statistics are rubbish’ takes are exaggerated. So we’ll give you a guide to the different indicators for the labour market, what they measure, and how they read for the UK.

The data sources

There are four main data sources for current labour market conditions in the UK:

• The Labour Force Survey (LFS) is an ONS-conducted study that samples 40,000 households (around 100,000 people) every quarter, asking them questions about their employment status, whether they are looking for work, and if they are not working, why not.
• ‘Real-time information’ (RTI), which every month is collected through the HMRC-administered pay-as-you-earn (PAYE) income tax system. RTI is an extremely timely and high-quality measure of the income tax system, and a good proxy for data on employees. It does not capture those self-employed, as they do not pay PAYE but instead self-assessment nine months after the end of the financial year; and it measures employments, rather than number of people in employment. That means someone with two jobs will appear twice in this data set.
• Workforce jobs (WFJ) is a quarterly, mixed-source measure created by the ONS, which takes into account LFS and administrative data, as well as employer surveys, and is used to produce a detailed industry-by-industry breakdown of employment.
• The claimant count is a tally of people claiming unemployment-related benefits from DWP – mainly universal credit (UC), although some legacy jobseeker’s allowance and other benefits remain. This is an administrative measure, and not an internationally agreed definition, and the introduction of UC has created comparability difficulties, which have led to its downgrading to an experimental official statistic.

The LFS is the headline release, which is meant to measure internationally agreed definition of employment, unemployment and inactivity. This is the release that was initially pulled and then withdrawn altogether, replaced by an experimental dataset based on growth from administrative sources.

The problems with the LFS stem from a falling response rate, which was as low as 15% in the latest survey, meaning that the ONS did not think it was representative. Response rates have been falling over the years, and with the survey no longer being conducted in-person since the pandemic, this has accelerated. This is not a UK-specific problem, but response rates seem to have suffered more in Britain.

The ONS has been aware of this issue, and has been testing a replacement survey – the transformed LFS, or TLFS – since late 2022. This survey has a bigger sample size and is done online, with in-person and telephone-based follow-ups, and has generally been getting higher response rates. It will be fully rolled out in the spring of 2024, but that means we face a few uncertain months as regards to labour market conditions.

This is far from ideal from policymakers’ perspectives, although in the short-run there seems to be little else that can be done other than look at a broader suite of indicators.

What is employment and how has it been moving?

Employment is the simplest of the concepts, as it is simply the number of people aged 16 and over who worked for at least an hour in a particular week (called the reference week) or were away temporarily from their job (e.g. on holiday). This is an internationally agreed definition by the International Labour Organisation (ILO), which the ONS uses.

The issue is that LFS-measured employment has remained much more subdued than that indicated by administrative data and other data.

Chart: UK measures of employment since December 2020

The chart above conveys the differences in measurement to some extent, but this becomes even more visible when looking at growth rates since December 2020.

Chart: Growth in UK employment measures since December 2020

Closely linked to this is the employment rate. There are two main measures, one looking at those of working age (16-64 in the ILO definition) and looking at all those over 16. Looking at those 65 and over is helpful when thinking about population ageing, but there shouldn’t be much difference in movements in the two in the short run.

Chart: UK employment rate since 2018

Much has been made in the last couple of years of the fact that the UK’s employment rate seems to be down considerably since the pandemic. But income tax receipts have performed well, and employment data from other sources painted a different picture – which the ONS has now attempted to adjust for in the experimental LFS statistics.

What we don’t know yet is what is happening to self-employment, and how that interacts with overall employment. It is possible that there could be some movement across from self-employment to employment, which would also be consistent with stronger growth in RTI data relative to overall employment. Administrative data is less helpful there – the self-employment don’t file tax returns until the January after the end of the financial year. So we will have to wait to know more.

Measuring unemployment is even trickier

Administrative data is a great help on the employment side, and especially for employees. But while there some unemployment-related administrative datasets, they are less directly related with headline measures of labour market participation and unemployment.

The main administrative dataset on unemployment is the claimant count, which DWP produce based on UC and other legacy benefits. But as the chart below shows, the relationship between the claimant count and LFS unemployment has fluctuated wildly over time.

Chart: UK LFS-based and claimant count unemployment

Why is there such a discrepancy? There are two reasons. One is that they are measuring very different things. LFS unemployment is an ILO-based definition, which counts people:

• without a job, who have been actively seeking work in the past four weeks and are available to start work in the next two weeks; or
• out of work, who have found a job and are waiting to start it in the next two weeks.

Claiming unemployment-based benefits is a much different process, and people might be unemployed but not claim related benefits for many reasons: some are UK-specific, others more general.

UK unemployment-related benefits are quite low by international standards, and do not directly relate to income replacement, which discourages some people from claiming altogether. This varies by country – in some cases, income replacement is standard and can last for a long time. There is also quite a bit of conditionality attached in the UK, including timelines for getting a job, which discourages claimants. This means the link between the two measures is weaker in the UK than in some other countries.

Apart from that, the economic cycle plays a part. In a robust labour market and with a positive economic outlook, people who are unemployed are less likely to claim as unemployment is more likely to be short-lived. On the other hand, during recessions or periods of slow growth, people are more likely to put themselves through the process of claiming.

There have also been reforms to the system which have changed likelihood of acceptance and generosity of benefits, such as the introduction of UC, which have changed the relationship between LFS unemployment and the claimant count. And pandemic support measures increased eligibility temporarily, which accounts for claims above the LFS-measured employment since 2020.

This means that little can be inferred directly from the claimant count about unemployment. We don’t know precisely the relationship between the two series, and there are lingering pandemic support effects in the claimant count. So there is no good substitute for being able to survey the population to estimate unemployment.

What does the current state of the labour market statistics mean for the UK and for Scotland?

Clearly it was right for the ONS to withdraw statistics it didn’t feel were representative, and it has attempted to give its most informed possible view of the labour market in the new experimental series.

But there remains considerable uncertainty regarding how much we can know about the state of the labour market. Unfortunately, this is a difficult juncture for the Bank of England and the Treasury, who would have been expecting to rely on labour market conditions to guide their policy interventions.

It also puts the Scottish Fiscal Commission and the Scottish Government in a difficult position for the Scottish Budget. The ONS has produced UK-wide experimental statistics, but not the usual regional ones based on the LFS. So we don’t have good estimates of employment and unemployment for Scotland. More weight than usual will have to be placed in RTI statistics for the SFC’s forecasts – but in particular the lack of certainty about the self-employed population will make Scottish Income Tax forecasts even trickier.

Over the longer run, the signs from the transformed LFS are positive, and we hope that higher response rates and a larger sample will contribute to more robust statistics for such an important part of the economy.