What better benefits data means for poverty in the UK

The Department of Work and Pensions (DWP) have improved the income data we use to measure poverty in the UK by linking administrative data about the social security benefits that people receive to their survey responses in the Family Resources Survey (FRS). They have done this for the first time with the 2024/25 data but have also looked back and retrospectively linked people’s benefit records from 2021/22.

While this helps us to get a better picture of people’s incomes in the UK, it means that there is a break in the series so we can’t compare the data from before this change in method, to after. This restricts what we can and can’t say about incomes and poverty in the UK, particularly if we want to talk about change over time. Here we explain:

• what has changed
• what challenges this raises
• how big a difference this has made to poverty measures
• what we can and can’t say about poverty.

Changes in how incomes are calculated

The DWP have collected data through a large household survey called the Family Resources Survey (FRS) since the mid-1990s. It is the most detailed survey of household incomes in the UK and collects information on income sources, amount of income, family characteristics, housing costs and other important questions about how households in the UK are managing financially.

What makes this such a good dataset to understand incomes and poverty in the UK is that it asks about income from lots of different sources like earnings, benefits and savings interest. This information is then used to build up the Households Below Average Income (HBAI) dataset which we use to measure and understand poverty in the UK. 

Respondents to the survey are asked whether they received any social security benefits and if so, how much. In many cases respondents might not know the exact amount of benefit they receive, or might not know or want to say that they receive a benefit. More rarely, someone might report that they are receiving a benefit they no longer are in receipt of. DWP do a lot of work to correct benefit records looking at respondents’ eligibility and standard amounts. Nevertheless, this has resulted in an undercounting of benefits in the FRS, both in terms of the number of recipients and the amount recorded.

People who are in receipt of a benefit will have information about this benefit saved within their DWP records. This includes what type of benefit they get and how much. For example, it will say whether they receive Universal Credit and if so, how much they get and for what, for example, whether they get the housing element.

The new method means that the DWP don’t have to rely on people reporting that they receive a benefit or how much they receive. Instead, with consent, they link their records directly to the people who took part in the survey which means they now know the benefits they receive and the exact amount.

This is much more accurate for a few reasons. It doesn’t rely on people’s memory. With a benefits system which is complex and sometimes confusing, it can be hard for respondents to remember the level of detail needed for this type of survey.

These questions are detailed and require individuals to have some understanding of the benefits that they receive, this will not always be the case, particularly for benefits like Universal Credit where there are different component parts that form 1 single payment to the individual. In addition to this, there is significant stigma around benefits in the UK. This can make people reluctant to share this information with interviewers. We also know that this stigma is not always shared equally across communities which means that under-reporting of benefit take-up/amount might be more prevalent in some communities than others.

An additional future positive to the new approach is that these questions take a long time to answer when people are asked in the form of a traditional survey, so using the administrative data to automatically provide answers could shorten the interview length (note at the moment, respondents are still asked for the information — it’s just that this is replaced by the linked administrative data).

The DWP have used linked administrative data for their main statistics for the first time with the 2024/25 data but have also looked back and retrospectively linked people’s benefit records from 2021/22. This means that between 1994/95 and 2020/21 we are relying on people self-reporting whether they receive benefits, and the amount they receive. From 2021/22 to 2024/25 we have values for self-reporting and the new linked data, with the administrative data now being used in the official statistics for all these years.

How does this change who is in poverty?

As we have explained above, at the moment, we only have this new data back to 2021/22 although there are plans to release 2018/19 through to 2020/21 linked data in summer 2026. This means there is a break in the series because of this methodological change, making comparison over time more challenging.

Although this linked benefits data is a big improvement to the data quality, there are a few challenges that this will cause in measuring and understanding how poverty has changed in the UK.

How it changes relative poverty

Relative poverty after housing costs (AHC) is the measure that is most commonly used to understand poverty in the UK. Someone is in poverty (AHC) when their household income, after they have paid their housing costs, is below 60% of the median, adjusted for family size and composition. Relative poverty before housing costs (BHC) is measured in a similar way but it does not remove housing costs. Relative poverty measures capture whether the incomes of poorer households are catching up with average incomes.

Benefit income was less accurately reported using the old method so the new data has:

• more households in receipt of benefits
• more household income from benefits.

This does not reflect a change in the actual take-up of benefits, or the actual incomes of households, but in how well they are captured in this survey. Many benefits (especially those more likely to be under-reported) are means tested or less likely to be received by high-income households, such as disability benefits. This means that the improvement to the data will have a bigger impact on the accuracy of incomes at the lower end of the income distribution than at higher levels.

This means that the recorded incomes of some households will have gone up, and this is more likely to have happened at the lower end of the income distribution by increasing the amount of income that is recorded as coming from benefits.

We see that there has been a slight change in the UK median income and, in turn, the poverty line which sits at 60% of the median income. In the old 2023/24 data, the median equivalised net household income (AHC) in the UK was £562 (on a couple without children basis). For the same year, using the new linked data, the median equivalised net household income in the UK is £582.

In turn, because the median income has changed using the new method, the poverty line (both before and after housing costs), has also changed. However, this change in the median is a relatively small contributing factor to why poverty rates from 2023/24 look different when using the old and new data.

The main reason that poverty rates look different is that many lower-income households that have seen an increase in their recorded income using the new method have been pushed further up the income distribution. At the same time, higher-income households are more likely to have stayed put. With the median income and, in turn, the poverty line remaining similar to where it was in the old data, if a household’s income has been raised high enough by the better recording of benefits data, then they can potentially move over the new poverty line.

The combination of a shifted income distribution and a changed median income means that poverty rates are different using the 2 different data collection methods.

As mentioned earlier, some household types might be more likely to under report the benefits that they receive. This means that the new linked data will better capture incomes for these groups too, but may also mean we see bigger changes in poverty rates for some groups compared to others.

How it changes absolute poverty

Absolute poverty compares people’s incomes to a fixed poverty line. Previously, absolute poverty was based on an inflation-adjusted 2010/11 poverty line (set at 60% of median income after housing costs in 2010/11). Absolute poverty rates tell us whether the incomes of poorer households are increasing faster than inflation. When the incomes of lower-income households are growing faster than inflation, we would expect absolute poverty to fall. This is important as it helps us to understand who the economy is working for.

The absolute poverty line was measured against the 2010/11 poverty line, partly due to the introduction of the Child Poverty Act 2010 (later largely repealed in 2016). It was designed to help government understand progress since the Act’s introduction. It was not selected for any analytical reason and now that the act has been repealed, there is no specific reason for using 2010/11 other than keeping it constant and comparable. It is also good practice to rebase the absolute poverty line periodically (before being based on the 2010/11 poverty line, the 1998/99 poverty line was used) to ensure the income threshold remains meaningful.

As mentioned above, we only have data using the new data collection method back to 2021/22. This causes a problem for continuing to use absolute poverty measured against the 2010/11 poverty line. We can’t compare a household’s income in 2024/25 (on the new basis) to the poverty line in 2010/11 (on the old basis).

This means that we need to rebase the absolute poverty line to a year where we have data collected in the new way. DWP have decided to use the latest year, 2024/25, as the new poverty line that absolute poverty will be compared against from now onwards. One justification for this is that the Government introduced their Child Poverty Strategy in 2024/25.

What does this look like for the old and new series?

Figure 1 illustrates the gap in the old and new times series and the impact of better data collection on the recorded poverty rate for all individuals, children, working-age adults and pensioners. As mentioned above, this difference doesn’t represent a true ‘fall’ in poverty, it represents a more accurate measurement of incomes in the UK. The amount of change in the poverty rate using the old and new data is greater for some groups than others. This is because households with children and pensioners are more likely to be in receipt of a low-income benefits or other state support, which means their incomes are more likely to have changed due to the new data collection method.

When talking about the difference in the poverty rates calculated using the old and new data, we have avoided using terms like ‘lower’ as this would suggest we are making a direct comparison between the 2 series which cannot be done as incomes are collected differently. It is clear that poverty rates have been revised downwards using the more accurate data. Yet this revision downwards is not the outcome of any policy changes or general improvements in economic conditions, just improvements to the data.

What does this mean for people on low incomes in the UK?

While the improved data collection method has resulted in revised income distributions, poverty lines and poverty rates between 2021/22 and 2023/24, the reality for people living in poverty, and around the poverty line, remains one where people are struggling to make ends meet.

The cost of living crisis continues to bite, affecting people both in and out of work. Struggling to afford basics such as the weekly shop is something ever-increasing numbers of people can relate to.

Grassroots Poverty Action Group

Improving the data is a positive step in understanding poverty in the UK, but it is not an end in itself. People living on a low income in the UK continue to struggle to afford the essentials, food bank use continues to be at close to record levels, and people continue to experience the stigma and trauma of living in poverty.

What can we say about poverty?

These improvements in the data will help us to get a better understanding of how families are managing now and what can be done to reduce poverty in the UK. But the break in the series does, and will continue, to make it harder to talk about how poverty has changed over time, and could lead to people using statistics in a misleading way. We have pulled together a quick set of prompts to help you to decide what you can and can’t say about poverty using the latest data.

What we can do and say:

• 1 in 5 people in the UK were living in poverty in 2024/25
• compare the 2024/25 poverty rate with the 2021/22 poverty rate using the new measure — for example: the poverty rate has stayed stable between 2021/22 and 2024/25, at 20%
• make comparisons between groups using the new measure (or the old measure pre-2021/22) — for example, we can say that in 2024/25, 27% of children are in poverty and this is higher than the proportion of working-age adults in poverty (19%)
• changes will make social security policy impacts more easily seen in the data.

What we can’t do and say:

• make an exact comparison between the old and new trend lines, for example: poverty has fallen from 24% in 1994/95 to 20% in 2024/25
• switch between which data collection series you are using without indicating this either visually (such as showing a break in the series on a chart) and/or in writing
• you should use the new, better-quality data, where possible — you can still use the old data for 1994/95 to 2020/21, as it remains the best source of data on poverty for this period
• suggest this implies improved take-up of benefits as this improvement means that the survey data better reflects the admin data, but the admin data has remained the same.

We all need to be clear and cautious when using poverty analysis produced using HBAI prior to March 2026. JRF (and likely other organisations using poverty statistics in their work) will not retrospectively change values in previous published work that reports on poverty from 2021/22 to 2023/24. This means that this work will be based on values created using the old data collection method. Therefore, users will need to be cautious and clear if comparing between analysis/publications before and after March 2026.

What happens next?

There are further improvements planned for this data. The changes this time will focus on how survey data is scaled up (or grossed) to represent the whole population. This involves making sure the data sums correctly to a series of totals, so there are, for instance, the right number of children, the right number of private renters or the right number of people in the West Midlands in the published totals. These are known as ‘control totals’. Two separate changes are planned:

1. The latest census data (conducted in 2021 or 2022) will be used to build up the new population control totals. This change is planned for the 2027 publication.
2. New control totals will be introduced, so the number of benefit recipients more closely matches totals from the DWP’s administrative records. This is needed because even after using the linked data, as has just happened, there remains an undercount. This is because the group of people responding to the survey doesn’t exactly correspond to the general population. The timing for this is more uncertain, but we would press for this to happen in 2027, so we can concentrate on assessing progress, rather than understanding changes to the data.

The latter change is likely to act in a similar direction to that caused by the data linking this year, reducing poverty levels further. But we do not know the likely scale of this subsequent change on poverty rates.