Are fewer people living in poverty than previously thought?

Today, the Department for Work and Pensions (DWP) released the latest official statistics on UK household incomes, covering the financial year 2024–25. Alongside measures of the income distribution and income inequality, these include the headline poverty figures against which the UK, Welsh and Scottish governments judge the progress of their child poverty strategies.

These statistics are based upon an annual survey of around 16,000 households, which are used to create the Households Below Average Income (HBAI) data (despite the name, households across the income distribution are in fact included). This year, DWP has improved the measurement of benefit income in the data. Previously, families’ income from state benefits (such as child benefit, universal credit or the state pension) was measured only if they reported receiving the benefit when completing the survey, and in some cases benefit income went unreported. For the new release, DWP instead measures the benefit income of almost all survey respondents using their administrative records held by the government.1 Statistics for the years 2021–22 to 2023–24 have also been updated using the new methodology.2 This allows us to investigate recent trends using the updated data and to evaluate the impact of the revision itself.

In this comment, we consider the impact of the change to how benefits data are collected, and re-evaluate post-pandemic trends in household incomes and poverty based on the updated statistics.

Improving the measurement of household incomes

DWP have started using administrative data on benefit income because data based only on survey responses undercount benefits. For example, while government records show that 12% of families received universal credit in 2022–23, in the survey just 9% of families reported receiving it.3 This primarily affects the measurement of poorer households’ incomes, since they are more likely to receive benefits. There are two reasons for the undercount. First, some survey respondents who are in receipt of benefits fail to report this when surveyed. Second, true benefit recipients are under-represented in the survey, even after the sample is adjusted (‘reweighted’) so that it is representative of the overall population in terms of demographic variables such as age, sex, region and housing tenure. Today’s revision will resolve the first of these issues, by ensuring that respondents’ benefit income is included whenever DWP can observe them receiving it in their administrative records.4

For the latest year (2024–25), we only have income statistics that rely on administrative records for benefit incomes. But we have statistics for 2021–22 to 2023–24 that use both survey-based and administrative data for benefit incomes, so we can compare these to understand the impact of the switch to administrative benefits data.

Figure 1 shows (for the 2023–24 data) the percentage difference in household disposable income (income after deducting tax and adding benefits) at each income percentile when using administrative benefits data rather than survey-based benefits data. At all income levels, there is an increase in income when using administrative benefits data. The increase in income is more substantial for households towards the bottom of the income distribution, with a 9.8% increase at the 10^th percentile due to the revision. Percentage changes are smaller further up the distribution because these households are less likely to receive benefits, although the impact on median incomes (i.e. at the 50^th percentile) is still notable – a rise of 2.9%. As we show later, these changes serve to reduce measures of income poverty.

Integration of administrative benefits data is part of a wider plan by DWP to improve the quality of the data underlying the official income and poverty statistics, with many changes either planned for future releases or under exploration. Two important future changes relate to DWP’s reweighting process, which attempts to adjust for the fact that certain kinds of households are more likely to respond to the survey than others, in order to make the data more representative of the population. These changes are to make use of the most recent Census data5 and to give extra weight to benefit recipients to account for the fact that they are currently under-represented in the data. DWP is also exploring linking to other administrative sources, including for earnings, to further improve the quality of the income measure in the HBAI data.

We can therefore expect to see further revisions to official income and poverty statistics in the coming years. As we will show, today’s methodological changes result in slightly higher estimates of median incomes and lower estimates of poverty rates. But we do not know which direction the future changes will push in. It is plausible, for example, that the upweighting of benefit recipients will to some extent push estimated poverty rates back up and estimated median incomes back down. This means it is too early to tell whether true average incomes are in fact higher than previously thought, or whether the true poverty rate is lower than previously thought.

Household incomes

We now examine what the new statistics show about recent trends in average household incomes and poverty, and how the new methodology has affected estimates. DWP has only revised data back to 2021–22 based on the new methodology so far, meaning data from earlier years are not strictly comparable. We focus on recent trends since the methodological revision.

Figure 2 shows median (middle) household disposable income since 2002–03. Disposable income is the sum of a household’s income from various sources such as earnings, pensions, benefits and savings, less the taxes it pays. To account for differences in family size (and hence needs), incomes are equivalised (adjusted for size) and expressed in terms of amounts for a childless couple. Official statistics with benefit income based on survey responses are plotted with a dashed line, while official statistics based on the administrative data are plotted with a solid line. Median incomes rose from £35,900 to £37,500 (5%) in the latest year of the data (2023–24 to 2024–25) when measured before housing costs are deducted (BHC), and from £30,900 to £32,500 (5%) when measured after housing costs are deducted (AHC). This is a substantial and statistically significant increase.

This fast growth likely reflects two factors. First, data on earnings from tax records show there was strong real earnings growth from 2023–24 to 2024–25. Second, last year’s HBAI data showed falls in real earnings that were not borne out by tax data, likely due to chance in who happened to be in the HBAI sample that year. So it is possible that some of the observed large year-on-year increase in average incomes simply represents an undoing of that discrepancy. Despite the strong growth in the latest year, median BHC income was only 2% higher than it was in 2021–22 (averaging less than 1% growth per year, from £36,800 to £37,500).

On average over the three years for which we have measures from both the old and new methodologies, median BHC incomes were 3% higher (£900) when using administrative benefits data rather than using survey responses. The year-to-year changes in median incomes are broadly the same.

Figure 3 shows how incomes grew across the household income distribution between 2021–22 and 2024–25. Broadly speaking, growth in incomes has been similarly slow for poorer and richer households. The estimates suggest lower-middle-income households have fared better, at 1% income growth per year, while growth was more muted for those at the 25^th and 75^th percentiles. Differences across the distribution are small, however, and should not be over-interpreted. Rather, what the figure makes clear is that no part of the distribution has escaped the slow disposable income growth seen since the pandemic.

Poverty and income inequality

Figure 4 shows rates of relative AHC poverty – the share of households with income below 60% of median income after deducting housing costs. As in Figure 2, official statistics with benefit income based on survey responses are plotted with a dashed line, while official statistics based on the administrative data are plotted with a solid line.

In 2024–25, the official estimate of the relative poverty rate among the whole population has remained broadly stable, rising very slightly to 20%, but is slightly down on the 2021–22 figure. By contrast, child poverty fell very slightly between 2023–24 and 2024–25, and is now essentially unchanged from the point it was at around the end of the pandemic. None of the year-on-year changes between 2023–24 and 2024–25 is statistically significant.6 For pensioners, there was a more notable (but still not statistically significant) 2 percentage point rise in relative poverty in the latest year. However, because pensioner poverty fell between 2021–22 and 2023–24, it was still 1 percentage point lower in 2024–25 than in 2021–22.

From Figure 4, we can see also that the impact of using administrative benefits data on poverty measurement is to reduce measured poverty by around 2 percentage points (ppt) over 2021–22 to 2023–24.7 This is because, as we saw in Figure 1, using administrative benefits data leads to a substantial increase in income for lower-income households. The differences are bigger for children and pensioners than for working-age adults (2ppt, 3ppt and 1ppt respectively). This is because children and pensioners are more likely to be in benefit-receiving households (pensioner benefits such as pension credit and the state pension are among the benefits counted) and hence in households whose benefit income is sometimes unreported when using survey-based benefits information. Measured trends differed only very slightly, with a slightly larger fall in poverty over the three years. But there are only three years of statistics calculated both with and without the methodological revision, meaning this is not necessarily evidence of any long-run systematic effect.

For the same reason that using administrative benefits data reduces measures of poverty – that it increases incomes much more for poorer households than for richer households – it also leads to lower measures of income inequality. The Gini coefficient measure of income inequality, a high-profile measure that takes account of incomes right across the distribution to give a score between 0 (complete equality) and 1 (all income held by one person), was 0.33 for 2024–25. Over the years where we have both measures, using administrative benefits data lowers the Gini by less than 0.01 (1 percentage point). The 90:10 ratio, another measure of income inequality which is defined as the 90^th percentile of income divided by the 10^th percentile of income, was 3.7 in 2024–25. Using administrative benefits data rather than survey-based data reduces the 90:10 ratio by 0.2 on average.

Conclusion

With the UK, Scottish and Welsh governments all tracking progress against official poverty rates, it is important that trends in poverty – and household incomes more broadly – can be estimated accurately. The improvement to benefit income measurement in today’s statistics addresses one long-standing weakness in the official data: benefit under-reporting. However, DWP is planning further changes to address other data issues, so poverty rates and other income statistics will be revised again in the coming years. We do not yet know how these changes will affect the estimates, and the figures may shift as they are introduced. Over time, though, they should lead to more robust measures of living standards.