Is welfare spending ‘out of control’?

Why this is a misleading claim, and how to understand the more complex reality...

Regular readers will know that many of my blogs are critiques of the ‘facts’ that underlie public debate – claims that we’re facing a crisis of worklessness because of health/disability, or that out-of-work benefit claims are much higher than they were a decade ago.

But perhaps the most important of these claimed political facts is that welfare spending is out of control.

In Rachel Reeves’ recent words, “It’s been 14 years of shameful failure and mismanagement by the previous government that has let welfare costs spiral out of control.” It’s perhaps unfair to single Labour out here; we saw the same rhetoric from the Conservatives (hence Rishi Sunak’s mooted cuts to PIP), and indeed, it’s partly shared by many respected analysts, such as the OBR, IFS and Resolution Foundation.

Still: I think that the ‘out of control’ claim is highly misleading – it contains a grain of truth, but used to build a house on sand, and take us away from the policies we need to fix the real issues. Let me explain why.

What has really happened to welfare spending?

We need to start by separating out spending on pensioner benefits, because – despite their importance – this is simply not what most people understand by ‘welfare’.¹ We can then see trends in welfare spending (as a share of GDP) in the figure below (which updates my recent report², and you can see a similar chart in the OBR’s 2022 Welfare Trends Report³).

This makes two things extremely clear:

1. Spending on all health/disability-related benefits has continued to rise – from 1.0→1.7→2.4% of GDP 1978/9→2010/11→2023/24. Within this, spending on disability benefits like PIP has risen most sharply, and is forecast to rise further.

2. Total welfare spending is similar to much of the last 45 years. In 2023/24 it’s expected to be 4.9% of GDP, which is lower than 12 of the 44 years that we have data for here.

This is why I get so infuriated when people present the figures on disability/incapacity benefits spending as a proxy for total welfare spending. You cannot understand the recent history of welfare without seeing declining spending on non-disability benefits, which have decreased from 4.0→2.6% of GDP 2009/10→2023/24. Obviously disability-related spending has risen, but this must be understood alongside what has happened to other welfare spending (indeed, the two things are causally related). But I’ll return later to what this means for policy.

Grains of truth

Now, it’s important to recognise the two grains of truth in the claim that welfare spending is out of control. Firstly, the times that welfare spending has been higher than now are associated with major recessions – the early-1980s, the mid-1990s, the late-2000s, and then the Covid spike. The level of spending at the moment is relatively high for a ‘normal’ non-crisis period (though we can debate whether the current period is actually ‘normal’, given the cost-of-living crisis). Secondly, the OBR are forecasting further rises in welfare spending, which again is unusual outside of a crisis (although the extent of the increases is relatively small if you take into account the cost-of-living payments ending). Still, given the pressures on the public finances, this means the spotlight will shine on welfare spending.

Yet even this more nuanced take might be overstating things. The rising state pension age (SPA) means that 625,000 people are claiming benefits in August 2023 who would not have been counted as ‘working-age’ back in 2010.⁴ This is 8% of the total number of claimants, so very crudely assuming that this is the % reduction of spending that they account for, welfare spending on a consistent age group (18-59/64 year olds) is probably ≈4.5% of GDP in 2023/24 (not 4.9%), which is not high historically. Moreover, presumably the upcoming OBR forecasts include the effects of a further 1 year rise in the SPA that happens from April 2026 to Feb 2028. It’s frustrating that this is bundled in with welfare being ‘out of control’, rather than part of a deliberate attempt to lengthen working lives (and save the Treasury spending on pensions!).

But shouldn’t we care about real-terms spending?

Before we get to the policy implications of all this, I need to deal with a potential objection: benefits spending may not be that high as a % of GDP, but it’s rising sharply in cash terms.

One version of this is pretty dumb. The Sun (in reporting of Rachel Reeves’ speech above) said “welfare costs have increased a staggering 64% since 2010 when the bill stood at £187 billion.” This is staggeringly wrong - it looks at the nominal spending in 2009/10 compared to 2024/25, without taking into account the 51% inflation that’s occurred since then.

(As an aside: I’m a bit worried that this might be a figure that the Treasury put out. If so, then it suggests they really want to distort the truth in order to get a ‘welfare is out of control’ story - which isn’t great. But I guess there’s a chance it’s just an error by the Sun in figuring out what some of the sums meant…).

But there’s also a real argument here, about what it means to say ‘welfare spending is rising’. The OBR often look at this in terms of real-terms spending - that is, spending after accounting for inflation. But what I’ve shown above is in terms of spending as a share of GDP - that is, spending compared to the total size of the economy. These differ enormously: real GDP (taking inflation into account) has risen by 140% since 1978, and 24% since 2010.⁵ So spending that is flat as a % of GDP is still a big real-terms spending rise.

Put simply, I think it doesn’t make sense to look at welfare spending in real terms. The point of the social security system is (among other things) to smooth people’s income between their times of need and times of plenty, and to protect against poverty. Both of these require benefits to keep up with wider living standards. Ever since reading John Hills’ timeless book Inequality and the State 20 years ago, I have understood that even the public’s conception of ‘absolute poverty’ changes over time because it’s tied to the country’s wealth.⁶ And if you are going to offer people meaningful protection if they lose their job, it needs to keep up with average incomes too.

But why did real-terms spending rise?

That said, the Conservatives froze benefits 2016-20, and otherwise benefits were only uprated by CPI. So it’s surprising to see that welfare spending is rising in real terms.

Partly - and usually missed - this is explained by an odd reason: CPI (used to update benefits with inflation) has been higher than the GDP deflator (used to create ‘real-terms’ spending figures) - see this IFS briefing.⁷ This has a surprisingly big effect: as far as I can tell, it artificially creates a 4.3% rise in real-terms welfare spending 2021-23, and a 7.8% rise 2019-23. (But there wasn’t any difference 2010-19). We should ignore this rise, which simply comes from the limitations of the GDP deflator as a measure; it doesn’t mean that claimants actually saw any more money in real terms.

But it’s clear that there are other reasons why spending has been rising since 2019 - whether on the adjusted real-terms measure, or (much more clearly) the GDP measure. It’s still not particularly high in historical terms (when we take out the effects of the rising pension age). But - crucially - there’s been a clear rise in spending on disability-related benefits, which is counterbalanced by a fall in spending on other benefits. And this is something that we need to address.

A pause

However, the policy implications of all this need to wait for next week - I got sidetracked by a deep dive into inflation and uprating, and this post is already long enough! To be continued… (But do comment/email in the meantime with anything that relates to this post).

1. Public definitions of ‘welfare’ are not necessarily the right ones, but it’s useful to present figures that respond to the way that people think. Obviously pensioner benefits are an important issue in their own right – for an excellent briefing on this, see the recent IFS presentation here – and have been the target of the first major benefits cut under Labour, the winter fuel payment.

2. This chart is similar to Figure 2 in my recent report After the WCA, but updated to correct one error (in children’s benefits, from 2003-4), and to include the cost-of-living payments that mostly go to welfare claimants (and are therefore probably most sensibly treated as part of the welfare budget).

3. The OBR chart divides benefits differently and also includes the Covid schemes (furlough and SEISS). But even accounting for this it shows slightly different figures (differing by -0.3 to +0.2% of GDP, without any trend), for reasons that are not clear to me. For example, even fairly straightforward categories like disability benefits (PIP, DLA) are higher in the OBR series than in mine. These are very clearly presented in the DWP Expenditure and Caseload tables (Table 4(i)), so it’s difficult to know how this difference arises - but usually the OBR charts are correct..

4. Figures taken from Stat-Xplore for benefit combinations by age. (I’ve previously noted that this series is a pain to construct, so let me know if it would be useful to see my working tables on this).

5. GDP figures are taken from the DWP Expenditure and Caseload tables for consistency with the figure above; GDP deflator figures are from Gov.uk; historic inflation data from ONS.

6. Obviously you didn’t need a mobile phone or internet access to be part of society 50 years ago, but you do now. There are other places that perhaps talk about this in more detail, but IATS is the one that influenced me the most. The most important bit reviewing public opinion data is p59-60, together with discussion p64-65, p209, p232, p234, p239.

7. See Figure 8 of the IFS briefing. The IFS explain the difference as being because the GDP deflator (used to adjust spending into ‘real-terms’) is focused domestically, while CPI - and people’s actual spending - is also affected by energy imports. They don’t give the figure that I use in the text, but they do say that this effect accounts for 30% of the 1.2 percentage point real-terms rise in welfare spending 2019-20 to 2024-25. (Some of the rest is due to pensioner spending, which is excluded here).