Overview

The UK study’s main models find that clinically vulnerable respondents experienced a larger post-2020 increase in sickness-related inactivity than comparable low-risk respondents: 1.3–2.1 percentage points, both significant at 5%. Against a pre-pandemic vulnerable-group prevalence of about 5–6%, that is roughly a 20–40% relative increase. The aligned-period estimates remain positive, but the result is relative to the comparison group—not a national total, causal decomposition, or identified mechanism.

Learning outcomes

Clinically vulnerable respondents’ post-2020 inactivity increase exceeded the low-risk group’s by 1.3–2.1 percentage points in the main models, both significant at 5%.
Against a 5–6% pre-pandemic vulnerable-group baseline, that relative difference is roughly 20–40%, making the group-level change substantively important.
The estimate is not the total UK effect: low-risk respondents were also exposed, and the study does not identify one mechanism or national share.

Keywords

UKinactivityvulnerabilitypandemicDiDbaseline

The Result That Holds: A Larger Rise in Sickness-Related Inactivity

The central result of the UK study is a relative one: after 2020, clinically vulnerable respondents became more likely than comparable low-risk respondents to report inactivity because of long-term sickness. This chapter focuses on the size and consistency of that difference, and on the judgment it supports. Readers should be able to describe the result in percentage points, understand why it is large relative to the vulnerable group’s baseline, and avoid turning it into a claim about the whole national rise.

A positive gap across the main comparisons

The study’s preferred Model 1 compares changes after 2020 between respondents classified as medium or high clinical risk and those classified as low risk. In the pooled pre/post specification, the estimated difference is 2.1 percentage points. In the two-way fixed-effects specification, it is 1.3 percentage points. Both estimates are statistically significant at the 5% level.

The two numbers answer the same broad relative question with different protections against confounding. The pooled model does not include individual fixed effects, so it remains more exposed to persistent differences between the risk groups. The fixed-effects model compares changes within observed respondents while absorbing characteristics that do not change over time, alongside year effects that capture shocks common to everyone. The fact that the fixed-effects estimate is smaller does not make the pooled estimate meaningless or prove that the fixed-effects model has removed every source of bias. It shows why specification matters when a result is reported.

A percentage-point estimate is a difference in the probability of being recorded in the outcome category, not a percentage of a person or a share of all cases. A 1.3-point estimate means that the post-pandemic change for the clinically vulnerable group exceeded the corresponding change for the low-risk group by 1.3 points under that specification. It does not mean that 1.3% of every vulnerable respondent left work, nor that 1.3% of the national population was newly inactive. Keeping “points” separate from “per cent” prevents a relative comparison from being inflated into a much larger claim.

Statistical significance answers yet another question. The 5% result indicates that the estimated difference is sufficiently distinguishable from zero under the model’s assumptions and reported uncertainty. It does not measure the social importance of the estimate, prove the identifying assumptions, or reveal the mechanism. The estimate’s substantive size comes from its percentage-point magnitude and its relation to the group’s baseline; its statistical support comes from uncertainty around the estimate. Both pieces are needed for a responsible reading.

The study also tests the pattern with an aligned-period comparison. Model 2 compares the within-window change in an earlier seven-year sequence, 2010–2016, with the corresponding change in a later sequence, 2017–2023. Its pooled estimate is 0.7 percentage points, and its fixed-effects estimate is 0.5 percentage points.

Published by DeepMox

FAQ

How large was the UK study’s main inactivity estimate?

It was 2.1 percentage points in pooled Model 1 and 1.3 points with fixed effects, both statistically significant at 5%.

What does the 20–40% figure mean for UK clinical vulnerability?

It is the approximate relative increase represented by the 1.3–2.1-point difference against a 5–6% pre-pandemic inactivity baseline among clinically vulnerable respondents.

Is the UK inactivity result the pandemic’s total effect?

No. It is the larger change for clinically vulnerable respondents relative to comparable low-risk respondents, who were also exposed.

Previous9 minUK Pandemic Comparison Methods and Limits

Next11 minWhy UK Benefits Do Not Mirror Inactivity

UK Clinically Vulnerable Inactivity Rise

The Result That Holds: A Larger Rise in Sickness-Related Inactivity

A positive gap across the main comparisons

FAQ