Written evidence from Civitas (CPM0002)
Introduction
This submission to the Work and Pensions Select Committee addresses the question of how best to measure poverty. There is a wealth of evidence linking poverty to worse social outcomes, which members of the committee will be familiar with already.
My thesis is that both standard and novel measures of poverty fail to provide an accurate measurement. Instead, simple reported measures of financial well-being prove more descriptive of genuine need, real poverty, as well as responsive to shifts in the economy.
Poverty is generally understood by the policy and academic establishment in a way that neither relates to what real people understand as poverty nor their economic circumstances. Specifically, I encourage the Committee not to endorse the Social Metrics Commission measurement of poverty in its current format.
The following argument pertains for the most part, to general poverty and not specifically child poverty. However, the lessons learnt from the general case are applicable to specific sections of the population.[1]
Absolute versus relative: concept and measurement
Attempts to measure poverty have shifted from direct observation and classification methods deployed by Seebohm Rowntree in his ground-breaking research in York, to survey methodologies that rely on a threshold specified by researchers. It is a shift from direct to remote observation.
A second shift has occurred whereby poverty is defined as insufficient resources below a fixed threshold to its being defined relative to a moving threshold. This is the difference between absolute and relative poverty. The seminal moment in this shift was Townsend’s book Poverty in the United Kingdom published in 1979.
It is a shift from having a bare-minimum to having what is typical.
There are two problems that arise from these shifts. The first is the problem of remoteness – statisticians removed from those they study have little idea what is sufficient to meet the needs of people they have never met. Those very same people, do know. Moreover, poverty can be observed directly since squalor seldom hides itself, if only you make the effort to go and see it. The statisticians, however, make a judgement based on numbers they themselves have not collected.
The second problem is that under the relative measure, ‘the poor are always with us’, since real gains in their standard of living of the poorest are masked due to the fact that real gains are also made by the median earner.
Indeed, that is the case when looking at the standard measures of poverty. Absolute poverty is classified as those with incomes below 60 per cent of median household income in 2010/11. Relative poverty is classified as those with incomes below 60 per cent of contemporary household income.
As seen in the graph below, and using data from the Institute for Fiscal Studies, child absolute poverty has apparently plummeted since the 1960s, while it has risen during the 1980s and held roughly steady since, on the relative measure.[2]
Figure 1. Child poverty rates (AHC) – source: IFS
However, the concept of relative poverty is one that takes precedence in our discourse on poverty, both in parliament, academic, and the media.
Problems
There are two problems associated with the relative measure and one associated with both relative and absolute measures.
The first is that relative poverty is at odds with how most people think of poverty.
Research by Stewart and Roberts (2018) analysed responses to a UK government consultation on child poverty measurement conducted in 2012/13. They analysed 251 responses from poverty specialists including charities, academics, activists, local authorities, politicians and service providers.
They write that “[r]esponses also overwhelmingly reflect a relative understanding of poverty” with only 12 out of 251 favouring an absolute measurement.
While this will be a self-selecting sample of expert opinion, it is of value in that these are the ones who will have had the most influence on government policy, since they are the ones who have responded to the call for evidence.
How does this elite understanding of poverty compare with what is understood as poverty among the British public?
The British Social Attitudes survey asks respondents what they understand the term ‘poverty’ to mean. Three different understandings of poverty are proposed:
“Would you say that someone in Britain was or was not in poverty if:
[a]…they had enough to buy the things they really needed, but not enough to buy the things most people take for granted?
[b]… they had enough to eat and live, but not enough to buy other things they needed?
[c]… they had not got enough to eat and live without getting into debt?”
Possible answers are that the respondent thought that person “was in poverty”, “was not in poverty” or “don’t know”.
The three items can be thought of as mapping onto a relative understanding of poverty [a], a generous absolute understanding of poverty[b], and a restricted absolute understanding of poverty [c].
Responses to these questions have been analysed by Hill (2004) who found in 2000, 27 per cent agreed with [a], 59 per cent [b] and 93 per cent [c] as definitions of being in poverty. Ties to political party allegiance, personal experience of poverty and income are found, although differences are small.
The surveys show that in contrast to the experts, people tend to favour the restricted absolute understanding of poverty.
In 2018 BSA survey, 88 per cent said someone was in poverty if they had not enough to eat and live without getting into debt compared to 28 per cent saying someone was in poverty if they could buy the things they really needed, but not enough to buy the things most people take for granted (Curtice et al., 2019).
It is clear that the view expressed by the experts differs wildly from most British adults.
While the experts favoured a relative measure, the people favoured an absolute one with the net balance amounting to +60 percentage points in favour of [c] over [a].
That the relative conception of poverty has little grounding among ordinary people is nothing new. Townsend himself concluded that, “[o]nly a tiny percentage of the [his] sample saw poverty as a condition relative to standards which were or were becoming widespread in contemporary society” (Townsend, 1979, p270).
The very same people Townsend studied, rejected the premises by which he was prepared to classify them as in poverty or not.
My own analysis of these data revealed the understanding of poverty as a relative phenomenon not to be tied to social class or education, but rather a political one – the more left-wing you are, the more likely you are to support it.
However, not even a majority of leftists support the relative measure.
One other way to look at this is to compare support for a relative understanding of poverty by newspaper readership. Support is more likely to be found but among left-leaning newspapers with Guardian readers among the most accepting (38%) and Daily Telegraph readers among the least (10%).
The second problem with the relative measure is methodological. During the financial crisis and the ensuing recession of the early 2010s, poverty on the relative measure appeared to fall. The reason for this counterintuitive finding is that in a recession, median income overall falls, and therefore there are fewer people who can be classified as falling below the 60 per cent threshold. As the economy reasserts itself, so too does median income rise and so poverty appears to rise.
All reports of a ‘rise in child poverty’ based on this statistic alone, are based on a misunderstanding of this statistical truism, and not evidence of an immiseration attributable to government austerity at the time.
To summarise, what is understood as poverty by the experts, is rejected as poverty by the average person, while the way relative poverty is measured fails to tell the story of what goes on in a recession.
The problem common to both absolute and relative measures is far greater in that they simply do not match with how people themselves describe their economic circumstances.
My own analysis of Understanding Society data shows that while 13 per cent are in relative poverty (before housing costs), the majority of them describe their financial situation as at least ‘getting by’ if not better. Indeed, there were more people classified as in poverty, following standard conventions, describing their financial situation as ‘living comfortably’ than ‘very difficult’.
Furthermore, 41 per cent of those classed as in poverty had access to satellite television at home, which implies a disposable income spent on non-essential items. This is hardly consistent with both the popular and elite understandings of ‘poverty’.
What is true of the relative poverty measure will be true of the absolute one too.
(Moreover, these statistics are based on income yet many people who are extremely rich may have no income since they can afford not to work and live off their wealth. All such people would be classified as in poverty on both absolute and relative measures. The extent of their contribution to the final statistics is an answerable empirical question.)
The Social Metrics Commission
The Social Metrics Commission is an alliance of poverty campaigners and experts that has produced a new measure of poverty.
It is a relative measure that incorporates income as well as measures of available resources such as housing, childcare, and disability-related costs-of-living. A poverty threshold of 55 per cent below a three-year average of total resources available is set.
The following criticisms can be levelled at the SMC measurement:
- The criticisms I made of the standard relative measure will apply equally – this is a political and elitist understanding of poverty detached from what everyone else understands as poverty, that will not match their own appraisal of their living standards.
- The SMC figure of 22 per cent in poverty is too high – compare with Eurostat data showing the share of people experiencing severe material deprivation at 5 per cent (2017).
- SMC poverty goes up as Eurostat severe deprivation goes down!
- Changes in SMC poverty are slight - between 2007 and 2009, it rises by 1.4 percentage points before falling again. Contrast this with the Eurostat deprivation measure more than doubling between 2008 and 2013.
To summarise, the anti-poverty experts have created a new measure that will not match either the understanding or experience of real people. Moreover, it fails to track other conceptually-related measures and thus cannot be said to be a valid measurement.
Unfortunately, the SMC measure cannot be replicated by independent researchers who do not have access to the right software. Its code is supplied in Stata format but is too complex and time consuming to be translated into other programming languages such as R or SPSS.
It is incumbent upon the SMC to show precisely what evidence they have for the validity of its new measurement if they want to convince the government or parliament to take this on as its official metric.
I first made these arguments in a blogpost for Freer.[3]
How to best measure poverty?
As a general thesis, statisticians are good at counting and measuring things, so long as the concepts in question are simple – height, temperature and so on.
When applying abstract nouns, such as poverty, to statistical data, this is much more difficult, as we have seen, especially since so much of it seems to be arbitrary or lie in the eye of the beholder.
The standard absolute and relative measures of poverty, as well as the Social Metrics Commission’s novel relative measure, all fail in that they are arbitrary classifications created by statisticians working remotely from those they study. To reiterate, they have little idea of what is required to meet the real needs of real individuals. That knowledge is best held by those individuals themselves, and therefore, it follows that the best way for statisticians to measure poverty is to ask people directly about their living standards.
Such measures exist, such as the Understanding Society measure of the percentage who report finding it quite or very difficult to get by financially.[4] (Moreover, statisticians should look at household consumption of key as well as luxury items as indicative of poverty).
None of the measures of poverty discussed above, are empirically sensitive to tell this story. This time-series dataset is also precisely what one would expect to see in a recession and note there is no prolonged immiseration after the crisis, casting doubt on so much of what is widely believed to have been the impact of government austerity.
Figure 2. Shares reporting financial difficulty – Understanding Society/UKHLS
Finally, we have two competing ways of measuring poverty – let us call them the statisticians’ (absolute, relative, SMC) and the people’s (reported financial wellbeing). The former are complicated, the latter simple. Ockham’s razor, assuming all things are equal between the two measurements, would compel us to accept the people’s measurement as the valid one since it is the simplest and therefore the most likely to be true.
Looking at the graph immediately above, during the current COVID-19 crisis, do we really want a measurement of poverty that is known to scarcely budge, when in the last crisis, twice as many people as before were saying they were finding things hard?
Other issues
Such poverty statistics are often misinterpreted as demonstrative of a group of individuals – the poor – who are always with us. That is true, but the reality is that the poor in these statistics are not the same people from year to year. From what analysis I have seen, those who are classed as poor in one year are seldom poor in subsequent years. Many with low incomes, will simply be young people beginning their careers who are replaced by other young people as they move up the ladder to higher incomes.
Looking at graphs that show the stability of relative poverty, it is easy to assume the state is required to intervene to ‘end poverty’ when such interventions would in practice mean ending all those little solutions to poverty on the part of the private individual, that are not observed in cross-sectional surveys. Accordingly, I would recommend the government to make analysis of panel data key to its measurement of poverty.
Any target to reduce poverty, should bear in mind this cycle of cohort replacement.
Child poverty itself, should be tied to the resources of the families in which they are living.
Why does it matter?
We need a measure sensitive to shifts in real need. The financial crisis saw a doubling of need but neither the standard relative or absolute measures or the SMC measure tell this story. They are unlikely to tell us the full extent of suffering that can be tied to the lockdown and pandemic.
Claims of one fifth or even “nearly half” of black people in poverty feed grievances which are ungrounded, leading to dismay and radicalism. Were such things to be true, then it would be a matter of national shame, but mercifully, they cannot be for the reasons I outlined above. This leads us to talk down the moral worth of our country.
The problem is we are over-estimating the share in poverty which encourages drastic political measures when actually, the economy works much better at meeting need than we give it credit for.
Bibliography
Curtice, J. C. (2019). British Social Attitudes: The 36th Report. London: The National Centre for Social Research.
Hill, J. (2004). Inequality and the state. Oxford: Oxford University Press.
Social Metrics Commission. (2018). A new measure of poverty for the UK. London: Social Metrics Commission.
Social Metrics Commission. (2019). Measuring poverty 2019: A report of the Social Metrics Commission. London: Social Metrics Commission.
Stewart, K., & Roberts, N. (2018). Child Poverty Measurement in the UK: Assessing Support for the Downgrading of Income-Based Poverty Measures. Social Indicators Research.
Townsend, P. (1979). Poverty in the United Kingdom. Harmondsworth: Penguin.
February 21