Benefits and hunger – beneath the headlines from the new FRS data (I)

On trends in claims, the £20 UC uplift and food insecurity, and why surveys under-count people claiming benefits

There are relatively few datasets that make the headlines year-after-year, but the Family Resources Survey (FRS) is one of them. It sets out the official statistics on poverty and incomes, and this week’s release saw lots of coverage about the rise in absolute poverty and deprivation from 2021/2 to 2022/3. I’m not going to go into the poverty figures, because the IFS have already done a fantastic job of explaining them. Instead, over a couple of blog posts I’m going to pick out a few other things that caught my eye from the new data that were easy to miss.

(The raw data aren’t available yet, but you can create bespoke tables on Stat-Xplore – data below are from Stat-Xplore if I don’t say which published table it comes from).

Benefits

Claims about ‘trends in out-of-work benefits’ in the media are wildly unreliable (as I’ve blogged about before), because the official statistics simply don’t count things consistently over time. Trying to put together a more consistent series, I’ve suggested that out-of-work benefit claims are probably lower now than they were a decade ago. But FRS allows us to go slightly further.

In the chart below, we can see trends in benefits as people’s main source of income. Note that this is different from ‘out-of-work benefit claims’ because (i) it includes in-work claims on UC, Tax Credits, PIP etc., and (ii) it only looks at people’s main source of income (and moreover, this is split between different benefits). Still, it’s a useful check on the claim that I made before. When the raw data is released, I’ll try to do (or persuade someone else to do!) a further analysis that’s more sharply focused on out-of-work claims…

What the new data shows, though, is that benefits as people’s main source of income is NOT spiralling out of control – it’s slightly higher than before Covid, but much lower than for most of the period after the 2008-9 financial crisis; only the 2017-18 to 2019-20 is lower. If we include disability benefits, then the current level is also lower than the whole of the 2000s.

(It’s completely unclear why disability benefits see such a sharp fall here as a main source of income – my guess is that there’s a change in the survey methods, but I haven’t been able to figure out what this is yet).

Hunger

There’s important headlines this week about rising food insecurity (see the IFS briefing). But it’s useful to see how this varies by benefit claims, shown in the table below.

What we see is that food insecurity fell among UC claimants in 2020-21 (when the £20/wk uplift was introduced), when it rose among ESA claimants (who didn’t get the uplift). Then in 2021-22, food insecurity rose among UC claimants (the uplift was removed in October 2021), when it fell among ESA claimants. But more recently, food insecurity just went up sharply among basically everyone (even though this is the first full year for UC claimants without the uplift).

Some of these trends might partly be because of the cost-of-living payments, but there’s other things going on here too – I’m doing various bits of work to get better estimates of the impact of the £20 uplift on food insecurity, so more on this later in the year.

Measuring benefits

Finally, a more technical issue, but one that is fascinating and important – the latest release includes a special report on why the number of benefits claims in the FRS survey is lower than the actual number of claimants (which we know from administrative data).

Perhaps the most valuable thing about FRS is that it links people’s survey responses to administrative data, so that it can check if they’re really claiming the things they say they are. There is extensive evidence that surveys miss out claimants, and the latest release shows this clearly. In the tables from the methodological report, they show that in 2022-23, the survey – a gold-standard, very expensive one – only captures 64% of UC claimants, 70% of ESA claimants and 80% of PIP claimants. (The equivalent numbers in 2018-19 were 71%, 69% and 80% respectively, and given that the UC caseload is different now, reporting behaviour probably hasn’t changed much).

What’s more fascinating though, is that they can explain why this is happening. Because they link the survey to administrative data, they can correct the survey responses to capture survey respondents’ actual claims. But when they do this, the survey still only captures 78% of UC claims, 81% of ESA claims and 92% of PIP claims. In other words, this helps, but it doesn’t fully explain why benefit claims are being under-reported in the survey. It’s worth flagging that the FRS response rates have fallen through the floor - from 60% in 2016 to 44% in 2020 to 25% since - but the situation is no better in 2018-19.

The conventional understanding is that survey under-reporting of benefits claims happens partly because of stigma, and partly because the system is so wildly complex. (As the methodological report says, “There has been a long-standing assumption that the FRS benefit undercount is likely to be due to under-reporting by FRS respondents.”). This new release confirms that reporting error is a part of the problem – but that even if we correct for this, under-reporting is considerable. This must mean that the survey sampling method is failing to capture the sorts of people that claim benefits – for the sorts of reasons that Dan Edmiston has recently talked about.

Finally, it sounds like the FRS team want to (i) drop the benefits income questions from the survey, and completely replace them with the admin data linkage, and (ii) it adjust the survey weights in future so that they adjust for benefits under-counts. These sound positive in some ways - it will make the estimates of people’s incomes more accurate - but they will also introduce major headaches for people trying to conduct analyses over time (because the new methods won’t be comparable). And fundamentally, I’m just slightly worried that they’re assuming that a new weighting regime will fix the problem of simply not reaching certain sorts of people…

In a follow-up post early next week, I’ll go into the data on disability and mental health trends in the FRS release.