Why has in-work poverty risen?

People living in working households are much less likely to be in poverty than those below pension age living in workless households. Yet over the past quarter-century, the relative poverty rate among working households has steadily risen – from 13.4% in 1994–95 to 18.4% in 2019–20. In the latest data (covering 2023–24), the in-work poverty rate stood at 18.0%. This comment summarises a new research paper, published in the Oxford Bulletin of Economics and Statistics, which explains why the in-work poverty rate increased in the quarter-century prior to the pandemic.

What are the trends in in-work poverty?

In our new paper, we study relative poverty, defined as being in a household with disposable income (after taxes and benefits and adjusted for household size) below 60% of median income in that year. We focus on incomes after deducting housing costs, for reasons argued here. For example, in 2023–24, a couple with two children with a net income of less than £28,430 per year after housing costs would be classed as being in relative poverty. Figure 1 shows the relative poverty rate among individuals in working households between 1994–95 and 2019–20. The poverty rate has risen by 5 percentage points over that period, from 13.4% to 18.4%.

Figure 1. Relative poverty rate for working-age families in a working household, after housing costs are deducted

_{Source: Figure 1 of Cribb, Waters and Xu (2025).}

What drove the rise in in-work poverty?

We use simulations to study what poverty rates would have been in 2019–20 if a specific part of the economic or policy environment had not changed since 1994–95. Specifically, we examine the role played by rising housing costs, growth in pensioner incomes, rising earnings inequality, and tax and benefit reforms. Together these factors explain almost 85% of the 5 percentage point rise in in-work poverty over the 25-year period. Those interested in methodological details should consult the journal paper; here we summarise our key results.

Housing costs

Two long-run shifts caused housing costs to rise by more for low-income households than for the rest of the population. First, low-income households are more likely to rent, and rents (net of housing benefit) in both the private and social sectors rose strongly between the mid 1990s and the late 2010s. In contrast, mortgage interest payments fell after 2008, cutting housing costs for higher-income households who were more likely to own their homes with mortgages, and thereby pushing up the poverty line based on median income after deducting housing costs. Second, the share of poorer working households in private rented accommodation, where average housing costs are higher, increased over time. This shift did not happen to the same extent among middle- and high-income households. As a result, the poorest and second-poorest fifth of working households saw average housing costs grow by 36% and 69% in real terms respectively over the 25-year period we examine, compared with just 3% among the highest-income fifth.

We find that had housing costs grown equally across the income distribution, in-work poverty would have been 1.8 percentage points lower in 2019–20, accounting for over one-third of the total rise.

Faster growth in pensioner incomes

The 25 years in the lead-up to the pandemic saw a marked rise in pensioner incomes, driven by rapid increases in state and private pensions. Growth in pensioner incomes was much higher than income growth for people below state pension age: median pensioner income (after housing costs) was 21% lower than median non-pensioner income in 1994–95, but only 3% lower by 2019–20. Naturally, this is good news for pensioners. But because relative poverty is defined with respect to median income of the population as a whole, rising pensioner incomes raise the poverty line and hence, if nothing else changes, the in-work poverty rate.

Had pensioner incomes grown at the same rate as non-pensioner incomes, growth in in-work poverty would have been 1.7 percentage points lower in 2019–20.

Earnings inequality

Earnings from employment among working households became more unequal over the period we study, reflecting a rise in male earnings inequality over much of the period (the largest source of household incomes) and an increase in assortative matching, whereby high-earning workers became more likely to partner with one another. Between the mid 1990s and the eve of the pandemic, real earnings rose by 16% at the 10^th percentile, 29% at the median and 45% at the 90^th percentile. Despite these large differences in household earnings growth, we find that even if earnings growth had been evenly distributed across working households, in-work poverty would only be 1.3 percentage points lower in 2019–20 than it in fact was. The effect is somewhat muted for three key reasons.

First, pre-tax household earnings and disposable (after-tax, after-benefit) household income that accounts for household size and structure – which the poverty statistics are based upon – do not line up perfectly. For example, only about half of those in the bottom fifth of the pre-tax household earnings distribution are also in the bottom fifth of the disposable income distribution; 13% of them are in the middle fifth. This means that boosting low-earning households’ pay also raises many middle-income households’ incomes, thereby lifting the median and raising the relative poverty line.

Second, people in the bottom fifth of the earnings distribution get about half their household income from employment, whereas for middle earners it is 96%. Even equal percentage growth in earnings across the distribution therefore adds proportionally less to total income at the bottom than it does to the middle. This reduces the extent to which more equal earnings growth can reduce in-work poverty.

Third, there are significant differences across the earnings distribution in ‘effective marginal tax rates’ – the rate at which taxes rise and benefits are withdrawn as earnings rise. Workers in the bottom tenth of the earnings distribution lose more than 50p of every extra pound they earn in higher taxes and (particularly) lower benefits. Conversely, those around the middle of the distribution lose about 30p. This further blunts the effect of pay rises on disposable income for low-earning workers relative to their middle-earning counterparts.

Taken together, these channels mean unequal earnings growth had a more modest effect on in-work poverty than one might expect.

Tax and benefit reforms

Finally, we study the impact of tax and benefit reforms on in-work poverty. There are three key sets of reforms that shaped the tax and benefit system over the period we study. First, tax credits were introduced in the early 2000s, and then expanded in the immediate wake of the 2008 financial crisis. These significantly increased state support for low-income households (both working and workless), especially those with children. Second, the 2010s witnessed a series of cuts to means-tested benefits. Third, at the same time during the 2010s, income tax liabilities were reduced, most prominently through big increases in the income tax personal allowance. These tax reductions were especially large for lower-income individuals – but this does not necessarily line up with low-income households. In fact, the effect on household incomes tended to be bigger for middle-income households, both because they are more likely to have two adults in work (and so gain twice from an income tax cut) and because low-income households often see a decline in benefits when their income tax liabilities are reduced.

We find that the effect of tax and benefit reforms between 1994–95 and 2010–11 – incorporating the introduction and expansion of tax credits – was to reduce in-work poverty by 2.5 percentage points; but that the reforms over the 2010s – when benefits and taxes were both cut – increased in-work poverty by 1.9 percentage points. This means that the cumulative effect of all reforms over the 25-year period from 1994–95 to 2019–20 was to reduce in-work poverty by 0.6 percentage points.

Conclusion

We take from these results several lessons for policymakers. First, because relative poverty rates depend not only on what happens to poorer households, but also on the experience of middle-income households, it is important to understand how the trends for these two groups interact. For example, policymakers might be a little more relaxed about the rise in in-work poverty once they know that a significant share of this is explained by big increases in pensioner incomes, which pushed up the relative poverty line. Second, a combination of high effective tax rates and differences between low pay and low household income mean that policies targeted at raising the lowest rates of pay, such as the minimum wage, may not have a large effect on in-work poverty. Third, housing costs have been an important contributor to rising in-work poverty, and so action to reduce those costs (such as through facilitating greater housebuilding) could prove an effective way of getting in-work poverty rates down.