Households and worklessness since 1997: what the figures are really saying

By Declan Gaffney

Confirmation bias is the tendency to find what you are looking for in the evidence while ignoring any facts which would require you to revise your pre-conceived views. Last week, the ONS’s annual bulletin Work and worklessness among households was released.These are essential materials for anyone interested in how employment is related to welfare, income distribution and poverty. But never mind all that: this is what Chris Grayling got out of the report:

“These figures are a further indictment of how the current system is failing families and is [sic] a shocking reflection of the scale of worklessness across the UK that this government has inherited. Some areas of Britain are suffering from intergenerational worklessness, which is why we must act now to ensure that children living in workless households are not left behind like their parents have been.”

Let’s get the cheap shots out of the way before looking at what the ONS release really tells us. Before raising the spectre of ‘the scale of worklessness…that this government has inherited’ Mr Grayling might at least have checked the scale of worklessness the previous government inherited from its Conservative predecessors. In May-June 2010, 19.2% of UK households were workless; in May-June 1997, the figure was 19.8%. As we shall see, this kind of crude comparison is essentially meaningless, but it does raise the question of why Mr Grayling of all people thinks he has a right to be shocked by these figures. At the same time, his assertion that ‘some areas of Britain are suffering from intergenerational worklessness’ is revealing of the ideological blinkers through which he is reading the data. The report is based on survey rather than panel data and therefore says nothing whatsoever about intergenerational patterns, to which the right tends to give exaggerated importance when it comes to explaining worklessness and poverty. This is confirmation bias on a grand scale- and not for the first time on these issues.

Given this misinterpretation, it is worth looking at what the ONS data actually is about. It differs from standard labour market statistics in that it concerns employment outcomes at the household rather than the individual level. There are a number of reasons for wanting to look at employment in this way. We use data on household incomes for the purposes of measuring relative living standards and poverty, so if we want to understand the relationship between employment and living standards, we need to look at employment at the household level as well. At the same time employment is not randomly distributed across households: there are more households in which nobody is working, and more households in which more than one person is working, than would be the case if employment chances were completely unrelated to household structure. (For a rigorous account, see Gregg and Wadsworth’s paper here). Both household structure and the distribution of employment across different types of household have changed over time, so understanding trends in employment and the income distribution requires us to unpick the impacts of these two types of change. The statistical evidence on household/employment patterns is therefore highly relevant to important aspects of the income distribution, labour markets and relative life chances – it deserves better than to be used as a prop for off-the-peg ideological soundbites.

So what does last week’s publication tell us about trends in household worklessness – the percentage of working age households in which nobody is in employment? The chart shows how the rate of worklessness changed for different household types between 1997-2008 and 2008-2010 – roughly, before and since the financial markets crisis. In the first period, worklessness rates fell across household types, but particularly among lone parent households. After May-June 2008, rates rose for all household types except lone parents. However it is only among single-person households that the rate rose more during the recession than it had fallen in the previous period. For other household types rates of worklessness are lower now than they were in 1997, particularly for lone parent and ‘other’ households with children. The fact that rates for families with children are lower now than in 1997 is on balance good news, as it may help mitigate the impact of the recession on child poverty [but see this from ippr]. (In the spirit of resisting confirmation bias from all parts of the political spectrum, the slight fall in worklessness among lone parent households since 2008 should not be read as evidence of the effectiveness of Labour’s third-term welfare reforms.)

Despite the reduced incidence of worklessness for most family types the fall in the overall worklessness rate from 1997 is modest (from 19.8% to 19.2%) and the number of workless households is higher by 208,000 than in 1997. Clearly there have been other developments: the overall numbers and rates reflect changes in the total number of households over the period and in the relative shares of different household types as well as the changes shown in the chart, and these developments are generally pushing in the opposite direction, increasing rather than reducing the number of workless households. The tables below break the change in overall numbers down into the effects of these changes for both periods and for 1997-2010 as a whole. They show the effect of (a) growth in the total number of households holding the shares of different household types constant, of (b) changes in the relative shares of the different household types (composition) and (c) changes in the rates of worklessness within each household type. The latter (in bold in the tables) allows us to assess the impact of labour market outcomes, which is what is relevant if Labour’s record in office is to be the issue. Negative numbers show that the component in question reduces worklessness.

Looking at the first table, the increase in the number of households between 1997 and 2008 (row A) alone meant that the number of workless households was 256,000 higher than it would have been other things being equal, while changes in composition (row B) added a further 147,000 to the number of workless households, with the relative growth of single person households having a particularly strong effect (left hand column, +128,000). The reduction in workless households due to improved labour market outcomes over this period` (row C) was -586,000, compared to the -183,000 suggested by the aggregate data (row D). The point here is that changes taking place outside the labour market have had a big impact on overall numbers, offsetting substantial improvements in labour market effects. For the period 2008 to 2010, unsurprisingly, changes in rates are the overwhelming drivers of increases in household worklessness except for lone parent households. Overall, worsening labour market outcomes account for +330,000 out of a total increase since 2008 of 389,000. The former figure should be compared with the -585,000 figure for the period 1997-2008, showing that the improvements between 1997 and 2008 have not been completely eroded by the rise in unemployment since 2008.

Over the entire period 1997-2010, changes in number and composition rather than worklessness rates account for the entire increase in the number of workless households. Nonetheless it is striking that even with substantial increases in numbers for most household types, the numbers of workless households are lower except for single person households. In fact the overall increase is entirely due to higher numbers of single person households, independently of the increase in worklessness in this household type over the period. Among other household types, even with the impact of recession and increased numbers, the number of workless households is 156,000 lower than in 1997.

Mr Grayling’s misreading of ONS’s data is consistent with the coalition’s routine communications strategy for these issues: ignore the impact of recession, insist that there were no improvements since 1997, and divert attention away from what is happening in the labour market to imputed family-level processes and ‘welfare dependency’. A government genuinely concerned with worklessness and poverty might be better off asking what it was that went right between 1997 and 2008 rather than using selective data to underpin an ideological narrative of Labour failure. This is not to suggest that everything improved between 1997 and 2008, that more could not have been done or that government policy was the only or main driver of better outcomes. The record is mixed here as in other areas. At the same time, we can’t ignore the fact that Labour prioritised worklessness among families with children during its time in office in a way that predecessor governments of all parties had never attempted. As the spending review approaches DWP ministers in particular should be looking at what expenditure cuts most risk undermining this priority.

(Of course there is no guarantee that this piece is free of confirmation bias. The data is available in Excel format here for anyone who wants to carry out their own analysis.)