New data suggest dramatic decline in living standards, but it may just be the quality of the data that has declined

Today, the Department for Work and Pensions (DWP) released the latest official statistics on UK household incomes, covering the financial year 2023–24 (April 2023 to March 2024). These statistics paint a bleak picture for living standards over this period, showing falls in household incomes across the board compared with the previous year (2022–23), including particularly large falls for the poorest households. This is despite lower inflation, a recovery in the labour market, sizeable increases to benefits in April 2023 and the continuation of some temporary ‘cost of living’ support measures.

It is important to note that these statistics – based on the ‘Households Below Average Income’ data (HBAI; despite its name, it covers all households) – are based on a sample of households (around 17,000 in the most recent year, which is down from 25,000 in the previous year). This can cause the statistics to diverge from reality for three broad reasons. First, households are selected to be surveyed at random, and so by chance in some years a richer or poorer set of households might be selected. As part of the statistics, DWP releases confidence intervals to illustrate this uncertainty; as we show below, these confidence intervals in fact indicate that the large falls in income seen in the latest data could be down to random sampling rather than reflecting real changes. Second, those that respond to the survey may differ in difficult-to-measure ways from those that do not respond, and the group that responds could change over time and become less representative of the general population. Third, survey respondents might report their income inaccurately (and the degree to which this happens could change over time). The second and third effects are not captured by confidence intervals. We show below that there are reasons to think this year’s income growth estimates may understate the true growth in incomes.

How did measured incomes change across the distribution?

Figure 1 shows trends in median (middle) household income since 2002–03, measured after taxes and benefits, and adjusted to account for inflation. Incomes are ‘equivalised’ to account for household size, and the cash numbers shown are the equivalent for a couple with no children. The green line shows median household income before housing costs are deducted (BHC), while the yellow line shows income after deducting housing costs (AHC). The statistics show that in 2023–24 median income before deducting housing costs was 2.0% lower than in 2022–23. This follows a 0.4% fall the year prior. However, these falls were not statistically significant, meaning if incomes did not in fact fall, it is plausible we would see changes of this magnitude due to random sampling. As a result, real median income in 2023–24 was 3.6% below the level seen in 2019–20 and 0.4% below its level in 2016–17.1 The figures for median income after deducting housing costs were similar, implying that there were not big changes in housing costs. The picture is largely the same when considering median household incomes among broad demographic groups – children, pensioners and working-age adults. These figures can be viewed by toggling the options in Figure 1.

Turning to how incomes have evolved across the income distribution, Figure 2 shows average annual growth in household incomes in the HBAI data for each percentile of the household income distribution. At every income level, household incomes measured before deducting housing costs fell between 2022–23 and 2023–24. These falls were particularly large for the poorest: the 15^th percentile of BHC income (the level of income which 15% of households have income less than) fell by 2.3% while the 10^th percentile of income fell by 5.7%. Taken together with changes in the previous three years, this decline implies that real BHC incomes in 2023–24 were lower at every point of the household income distribution than they were in 2019–20. These trends look similar when we consider AHC income, visible by toggling the option. The graphs for 2023–24 compared with 2022–23 also show the 95% confidence interval for income growth at each point in the distribution; importantly, for almost all percentiles the confidence interval includes zero, suggesting that it is possible that these falls in income, even the very large falls towards the bottom of the distribution in the latest year of the data, were a result of random sampling rather than actual large falls.

Examining the drivers of falls in income, we find that falls in net earnings are the largest single contributor (0.8 percentage points of the 2% fall in the mean). As we show below, changes in gross employee earnings measured in HBAI this year are not corroborated by other data sources, showing caution is needed in interpreting these results. Also, smaller falls in income from benefits, and other sources, were seen in the HBAI data.

How many people were in poverty?

Although the 2023–24 data show that household incomes fell particularly sharply among poorer households, measures of poverty were broadly stable. Figure 3 plots rates of poverty (based on incomes after deducting housing costs) since 2002–03, including options to show rates for various demographic groups. The measured rate of absolute poverty, where the poverty line is fixed (after adjusting for inflation) over time, was 18.3% – equivalent to 12.3 million people – in 2023–24 (barely up from 17.9% in 2022–23). The relative poverty rate, defined as having an income below 60% of median household income, was 21.1% – or 14.2 million people – in 2023–24 (down slightly from 21.4% in 2022–23). 

This might seem surprising given the dramatic income falls for poorer households shown above, but is explained by the fact that falls in household incomes were largest for the lowest-income 10% – already well below the poverty line. Around the poverty line, incomes only fell slightly (explaining the muted rise in absolute poverty), and indeed fell less than the median (explaining the slight fall in relative poverty). Splitting by demographic group shows that absolute child poverty and pensioner poverty rose by more (1.4ppt and 1.0ppt increases), whereas absolute poverty among working-age adults fell slightly (by 0.2ppt).

As well as income-based measures of poverty, HBAI also includes a measure called material deprivation, which gives the proportion of people unable to afford a number of essential goods. The data show that 11% of pensioners, 23% of working-age adults and 28% of children are counted as being in material deprivation. Due to a change in the methodology and data collected on material deprivation, resulting in an updated definition, these estimates are not comparable to previous years. 

Click the dropdown below to see more data on material deprivation, food security, and other indicators of hardship.

How do the HBAI estimates compare with other data sources?

The trends presented here, based on the official household incomes data (HBAI) for 2023–24, are somewhat surprising. While they could be an accurate representation of reality, as noted above these statistics are calculated based on a survey of a random sample of households, and so changes in estimates can also be driven by changes in the nature of the sample selected. The HBAI data are reweighted to make them representative of the population with respect to a number of characteristics, and this considerably reduces the risk of random sampling causing estimates to vary, but there is still a risk of the sample of respondents changing in ways that are not captured by the reweighting exercise; there is also a risk that the accuracy of respondents’ answers to the survey could change.

One particular cause for concern is the well-publicised post-pandemic decline in response rates to surveys in the UK. The response rate for the Family Resources Survey, which is used to create HBAI data, fell from 49% in 2019–20 to 31% in 2023–24 – which could increase the risk of the sample being less representative. There have also been changes in the method of interviewing during and since the COVID-19 period which could change the quality of income measurement.  

In this section, we look to alternative sources of data on incomes, to help us gauge how confident we ought to be in the trends the HBAI data show. HBAI is the best data source we have on overall household disposable incomes, but for some specific components of income more reliable measures exist, based on government records (often referred to as administrative data). There are some differences in what is measured between HBAI and the government data,2 but we have aligned things as closely as possible to make the comparisons informative.

Figure 5 shows employee earnings growth, after adjusting for inflation, by percentile of the earnings distribution, according to the HBAI statistics (solid) and administrative payroll data (dashed). The blue lines show earnings growth as measured in the two datasets between 2022–23 and 2023–24. For the majority of the earnings distribution, the earnings growth in the HBAI data is much lower than we observe in the administrative data. For example, median earnings in the HBAI data fell by 1.8% in real terms; in the administrative data, they grew by 0.3%. Notably, we do not see such large differences for other recent one-year periods. Given the important role played by falls in employment income in explaining the large drops in income estimated in HBAI, this discrepancy casts some doubt on the accuracy of these estimates.

A related issue is the number of employees observed in the data. In 2023–24, the number of employees in the HBAI data was 9% lower than the number implied by the administrative data, as shown in Figure 6. The gap was similar in 2022–23 – suggesting a limited role for driving differences in income growth between those two years – but in 2019–20 the gap was 7%. This is likely to push down income growth since the pandemic in the HBAI data relative to the true figure.

Figure 7 plots the total receipt of certain benefits estimated using the HBAI data, as a fraction of the total spending on those benefits according to government records. Prior to the pandemic, around 25% of means-tested working-age benefit income was missing in the HBAI statistics, and this was broadly stable. Since then, there has been a significant amount of volatility. By contrast, the undercount in working-age disability benefit expenditure (and the number of claimants) had been worsening over time, and has improved since the pandemic, whereas the undercount in pensioner benefits has been broadly stable over time. DWP analysis has identified that both undermeasurement of benefits among survey respondents and undersampling of those on benefits have contributed to an undercount in benefits in survey data. Importantly for comparing the latest year’s data with previous years, the share of expenditure on working-age means-tested benefits in HBAI in 2023–24 was about the same as in 2022–23 and 2019–20, but a higher proportion of disability benefit expenditure was captured in 2022–23 and 2023–24 than in 2019–20.

Conclusion

The HBAI data are the best source of information on disposable household income we have, and this year’s release shows a significant fall in incomes across the distribution, especially for the poorest households. But a key driver of this fall in income was a large fall in employee earnings, and this is not corroborated by data on tax records from HMRC, which should give a more accurate picture of earnings growth. This calls into question whether incomes did really fall so substantially, or indeed at all. Random variation between samples, falling survey response rates and changes to data collection methods may all have contributed to discrepancies between HBAI and other data sources. 

HBAI data are used to estimate average household incomes, inequality and poverty statistics, all of which matter greatly for informing policymakers. For example, the UK government is currently working on developing its child poverty strategy. Problems with the reliability of income data make it harder to precisely understand the nature of child poverty in the UK, and to set targets against which the success of government policies can be reliably measured. Improving the reliability of economic statistics will therefore be crucial in ensuring effective policymaking.