Working paper · Labour economics
Do disability benefits stop people working?
Most disability benefits shrink as earnings rise, so it is hard to tell whether they discourage work through the extra income or through the rules. The UK’s Personal Independence Payment has no earnings test, which lets us separate the two.
The Employment Effects of Disability Benefits Without Work Restrictions · Bouke Klein Teeselink, Georgios Melios · Working paper, latest version
Chapter 1
Two reasons a benefit might keep people out of work
Disability benefits are among the largest social insurance programmes in rich countries, often more than two percent of GDP. Most studies find they reduce employment among recipients, typically by 15 to 30 percentage points.
That fall can come from two places. An income effect: extra money makes work less necessary. A substitution effect: the benefit is cut when people earn, so working pays less. The first is a transfer that does not distort choices; the second creates real economic losses. Because most programmes do both at once, telling them apart has been hard.
If a benefit is never cut when people earn, any drop in work has to come from the income itself.
The reform that introduced PIP also changed who could get it, and where they were assessed. That created two natural experiments.
Chapter 2
Using the reform as a test
PIP widened eligibility for people with mental health conditions and narrowed it for people with minor physical disabilities. Because these groups were defined by health before the reform, we can compare how their working lives changed with similar people the reform treated differently.
The government also split assessments by region between two private providers. Atos covered eight regions, including London, the North and Scotland; Capita covered four, including Wales and Northern Ireland. If one provider was stricter, otherwise similar people faced different odds of getting the benefit simply because of where they lived.
Three comparisons
Gained access
Mental health problems
People with mental health problems before the reform, compared with similar people without them.
Lost access
Minor physical problems
People with minor physical limitations, compared with those with major ones.
Stricter screening
Assessor region
Regions assessed by Atos, the stricter provider, compared with regions assessed by Capita.
How the comparison is made fair
Each person in a treated group is matched with the five most similar people in the comparison group, on health, age, income, gender and education before the reform. We then compare their paths before and after 2013 (difference-in-differences), check that the groups moved together before the reform, and run placebo tests on earlier data.
The data come from a large household panel that interviews the same people every year.
Chapter 3
Following the same people for a decade
The UK Household Longitudinal Study interviews a representative sample of UK households every year. It records employment, every source of income including disability benefits, health and family circumstances.
We follow everyone aged 59 or younger in 2013, so that retirement does not drive the results, and use the years up to 2019 to keep the pandemic out of the main estimates.
First, did the reform change who received the benefit? Then, did that change who worked?
Chapter 4
Big changes in benefits, little change in work
The reform moved benefit receipt exactly as intended. The question is what happened to employment.
The question
Did better access to the benefit make people with mental health problems work less?
The answer
No.
Benefit receipt rose by 5.9 percentage points, but employment did not fall. The estimate is +0.3 points with controls (95% interval −2.2 to +2.8), and every post-reform estimate is slightly positive.
+5.9
points more likely to receive the benefit
The question
Did losing access push people with minor physical problems into work?
The answer
No.
Benefit receipt fell by 5.6 percentage points, but employment was unchanged: −0.3 points (95% interval −2.5 to +1.9).
−5.6
points less likely to receive the benefit
The question
Did stricter assessment raise employment?
The answer
Possibly, with caveats.
In regions with the stricter assessor, benefit receipt fell by 1.8 points and employment rose by 2.9 to 3.3 points. But the effect shrinks to +1.3 points and loses significance when people who moved between regions are excluded, and one pre-reform estimate is significant, so we read it as reduced-form evidence, not the effect of the benefit itself.
+3.3
points in employment, with controls
Employment, every comparison and specification
- No controls: 1.6 percentage points, 95% confidence interval −0.9 to 4.1, which includes zero.
- With controls: 0.3 percentage points, 95% confidence interval −2.2 to 2.8, which includes zero.
- No controls: −0.3 percentage points, 95% confidence interval −2.7 to 2.1, which includes zero.
- With controls: −0.3 percentage points, 95% confidence interval −2.5 to 1.9, which includes zero.
- No controls: 2.9 percentage points, 95% confidence interval 0.7 to 5.1.
- With controls: 3.3 percentage points, 95% confidence interval 1.1 to 5.5.
- No controls: 1.1 percentage points, 95% confidence interval −1.1 to 3.3, which includes zero.
- With controls: 1.3 percentage points, 95% confidence interval −0.9 to 3.5, which includes zero.
Matched difference-in-differences with individual and year fixed effects, UKHLS up to 2019. "With controls" adds age, age squared and number of children. 95% intervals computed from the reported standard errors, clustered by individual.
Show the numbers as a table
| Outcome | Estimate (pp) | 95% CI | Control mean | n |
|---|---|---|---|---|
| Mental health problems: more access: No controls | 1.6 | −0.9 to 4.1 | · | · |
| Mental health problems: more access: With controls | 0.3 | −2.2 to 2.8 | · | · |
| Minor physical problems: less access: No controls | −0.3 | −2.7 to 2.1 | · | · |
| Minor physical problems: less access: With controls | −0.3 | −2.5 to 1.9 | · | · |
| Stricter assessor region: No controls | 2.9 | 0.7 to 5.1 | · | · |
| Stricter assessor region: With controls | 3.3 | 1.1 to 5.5 | · | · |
| Stricter assessor region, excluding movers: No controls | 1.1 | −1.1 to 3.3 | · | · |
| Stricter assessor region, excluding movers: With controls | 1.3 | −0.9 to 3.5 | · | · |
First, the reform did change who received the benefit
- Mental health problems: 5.9 percentage points, 95% confidence interval 3.9 to 7.9.
- Minor physical problems: −5.6 percentage points, 95% confidence interval −7.4 to −3.8.
- Stricter assessor region: −1.8 percentage points, 95% confidence interval −3.4 to −0.2.
With controls. Table A17 of the paper. 95% intervals computed from the reported standard errors.
Show the numbers as a table
| Outcome | Estimate (pp) | 95% CI | Control mean | n |
|---|---|---|---|---|
| Mental health problems | 5.9 | 3.9 to 7.9 | · | · |
| Minor physical problems | −5.6 | −7.4 to −3.8 | · | · |
| Stricter assessor region | −1.8 | −3.4 to −0.2 | · | · |
Checks: 17 of 18 placebo estimates on pre-reform data are indistinguishable from zero. Of 51 estimates on other benefits, only three are significant, and benefits unrelated to disability policy do not move, which argues against austerity or other shocks driving the results. In stricter-assessor regions, the employment rise is concentrated among unmarried people (+6.0 points, against +1.5 for married people).
So what does this mean for how disability benefits are designed?
Chapter 5
Support without a work penalty can work
For people whose eligibility changed because of objectively measurable health conditions, gaining or losing PIP did not change whether they worked. That suggests a benefit designed to cover the extra costs of disability, rather than to replace earnings, can provide security without discouraging work.
The stricter-assessor results point to a second lesson, with more caution: tighter screening may help separate genuine need from claims by people who could work. The paper sets out the policy implications:
- 1Keep PIP’s work-neutral design, which lets recipients work as much as they can
- 2Consider widening coverage where conditions can be verified objectively
- 3Use rigorous assessment to improve targeting and limit strategic claims
The paper
The Employment Effects of Disability Benefits Without Work Restrictions
Bouke Klein Teeselink, Georgios Melios · Working paper, latest version
Funded by Ethos Lab as part of the Horizon Europe project BENEFITS (Grant Agreement No. 101179032). The results are research outputs and do not necessarily reflect the views of the European Commission or other organisations in the BENEFITS project.
Data: UK Household Longitudinal Study (Understanding Society), University of Essex, Institute for Social and Economic Research.
Working on a similar question?
Tell us what you need to know and by when. We will reply within two working days with an honest view of what we could find out.