The Gap Between Leaving Work and Getting Support
The subject of this series is the United Kingdom’s post-pandemic rise in health-related economic inactivity and the way that change did—or did not—appear in benefit receipt. This opening focuses on the gap between those two signals: what each one counts, why their divergence matters, and what evidence can legitimately be asked to explain. The practical judgment is simple but important: before treating a rising inactivity figure as evidence of a matching rise in claims, identify the stage of the process that the figure observes.
A sharp reversal in the UK labour market
The official numbers make the question hard to ignore. The Office for National Statistics series titled “LFS: Econ. inactivity reasons: Long-term Sick: UK: 16-64:%” reports a percentage, not a benefit caseload. In the longer record, the figure was 21.4% in 1993, rose through the 1990s to 27.0% in 1998, and then generally declined until the late 2010s. It stood at 23.7% in 2019. After the pandemic, the direction changed: 25.8% in 2020, 26.2% in 2021, 27.6% in 2022, 29.8% in 2023, 29.9% in 2024, and 30.7% in 2025.
That is a seven-percentage-point increase between 2019 and 2025 in the share of 16–64-year-olds recorded as economically inactive because of long-term sickness. It is not a small fluctuation around a stable trend. The IZA study that anchors this series reports that the number of working-age adults in this category had passed 2.8 million by early 2024, while also warning that problems affecting the Labour Force Survey make the exact size and timing of the change difficult to pin down. The warning does not erase the reversal. It tells us to be careful about what kind of conclusion the series can bear.
The distinction between a level and a flow is useful here. A percentage describes how many people are in a category at the time measured; it does not reveal, by itself, how many newly entered it, how long people remain there, or how many left it. A higher figure could therefore reflect several movements within the population rather than one single wave of new cases. The ONS series is valuable for showing that the category became more prevalent, but it is not designed to narrate each person’s route into or out of it.
There is a second scale issue. The ONS series is an official aggregate percentage for people aged 16–64, whereas the IZA study uses a linked longitudinal survey sample with its own age, observation, and outcome rules. Both illuminate the same broad UK problem, but they do not provide interchangeable denominators. A change in the official series cannot simply be divided by a subgroup estimate, and a subgroup estimate cannot be pasted onto the national trend as though it were a population share.
An aggregate percentage can establish that a population indicator moved. It cannot, by itself, tell us whether the people behind the movement applied for support, satisfied a benefit’s eligibility rules, completed an assessment, or received an award. Nor does it identify a single cause. Health may have deteriorated; treatment may have been delayed; people may have avoided workplaces; employers may have had fewer suitable jobs; or several of these things may have occurred together. The number is a signal of labour-market status, not a complete account of the route that produced it.
The same shock can produce different counters
The temptation to join the numbers is understandable. If more people are unable to participate in the labour market because of sickness, one might expect more people to claim disability or incapacity benefits. At a broad level, the outcomes are related: health and work capacity affect both. But they are not two names for the same event.
Economic inactivity is a classification of labour-market position and stated reason. It asks whether a person is participating in the labour force and, if not, whether long-term sickness or disability is the reported reason. The ONS series aggregates that kind of status across a defined age range. The longitudinal study uses a corresponding self-reported labour-market status outcome. Neither measure is an award decision.
A disability benefit asks a different question. Personal Independence Payment and its predecessor, Disability Living Allowance, are based on functional limitations and the additional costs associated with disability. A person does not have to be out of work to receive them. Someone may therefore remain employed and qualify for this kind of support, while someone who has left work may not apply, may not meet the functional test, or may still be waiting for a decision.
Incapacity-related support creates another route. It concerns limited capacity for work and interacts with programmes such as Employment and Support Allowance and the health-related elements of Universal Credit. The route can involve an application, evidence, a Work Capability Assessment, an administrative decision, and sometimes reconsideration or appeal. The timing of a recorded award can consequently differ from the timing of a health change or a labour-market exit. A person can also receive more than one kind of support, so the categories are not mutually exclusive boxes.
The distinction matters in both directions. A rise in sickness-related inactivity does not automatically imply a proportionate rise in claims. Conversely, a rise in claims does not prove that the same number of people have left work: disability support can be paid to people who are working, and caseloads can be affected by application behaviour and administrative processes. Treating the two counters as interchangeable produces a false choice between “the health problem is real” and “the claims are inflated.” The more defensible question is where, and how strongly, the two pathways connect.
The counters also run on different clocks. A person’s labour-market status can change at the point when work becomes unsustainable or no longer available. A benefit record may appear only after a person decides to apply, assembles evidence, is assessed, and receives a decision. The reverse can happen too: a functional-support award may be recorded while the person remains in work, because the award is responding to the extra costs or limitations of disability rather than to a complete withdrawal from the labour market. Comparing the dates or totals without accounting for those clocks can make a delayed relationship look absent or an overlapping relationship look identical.
This is why the phrase “benefit claims” needs more precision than it usually receives in public debate. A claim, an application, an assessment, an award, and ongoing receipt are not necessarily the same event. The evidence available here is strongest when it keeps the outcome labels intact: sickness-related inactivity for labour-market participation, disability benefits for functional limitations, and incapacity benefits for a separate work-capacity and programme process. The distinctions are not bureaucratic decoration. They determine which people can enter each count and when.
A relative question, not a national conversion rate
The IZA paper, The Impact of the Pandemic on Health-Related Inactivity and Benefit Claims, makes that question more precise. Its primary subject is not an abstract relationship between two statistics. It is the post-pandemic experience of working-age adults classified as clinically vulnerable to COVID-19, compared with otherwise comparable people classified as low risk. The study asks whether the rise in health-related inactivity and benefit receipt was disproportionately concentrated in the clinically vulnerable group.
The classification is based on pre-existing conditions associated with a higher risk of severe illness and with official advice that led some people to shield. Medium- and high-risk respondents form the comparison’s treatment group; low-risk respondents form its control group. That label should not be read as a diagnosis of infection or long COVID. It is a marker of differential exposure: these respondents were more likely to face serious health risks and stronger reasons to change their behaviour during the pandemic.
This also means that the study does not compare a pandemic-exposed population with people who escaped the pandemic. Low-risk respondents could become infected, experience disrupted healthcare, face the wider economic shock, or change their behaviour too. The study therefore asks whether the gap between the two risk groups widened after 2020. It does not provide a national conversion rate from additional inactive people to additional claimants, and it does not assign the whole UK increase to clinical vulnerability.
A simple hypothetical makes the estimand clearer. Suppose sickness-related inactivity rose in both groups, but rose more in the clinically vulnerable group. The study’s estimate would describe the extra widening of that gap, not the full increase experienced by either group. If the low-risk group also moved sharply, the relative estimate could be smaller even while the pandemic’s total labour-market effect was large. Conversely, a relative gap could widen for reasons that are not captured by a single measure of infection or illness. The comparison is informative precisely because it is narrower than the national story.
Clinical vulnerability is therefore best understood as a bundle of exposure and response. Pre-existing conditions may raise the risk of severe illness; official advice may encourage shielding; and the same conditions may affect capacity to work or recovery. These possibilities travel together in the classification. The study can ask whether the classified group’s outcomes changed differently, but it cannot use that label to decide which link was decisive. Keeping the bundle intact prevents a statistical comparison from becoming an overconfident medical explanation.
That boundary is not a technical footnote. It determines what a later result means. If clinically vulnerable respondents became more likely than low-risk respondents to report long-term-sickness inactivity, the evidence would show a relative deterioration in their labour-market participation. If the same comparison did not produce a robust difference in PIP, DLA, or incapacity receipt, that would not make the inactivity result disappear. It would indicate that the institutional path from reduced participation to a recorded award is selective, delayed, or governed by criteria that the inactivity measure does not use.
The first task, then, is not to decide which number is “the real one.” It is to understand what each number observes. Once the categories are kept separate, the apparent contradiction becomes an empirical question: what exactly is being counted when a person is recorded as inactive, receives a disability benefit, or receives incapacity-related support?