Does measuring poverty multidimensionally make a difference?
There have been various attempts in Australian research to measure the 'multidimensional' nature of poverty- that is, adding things like rental stress or health inequity to ordinary income measures. In this post, which originally appeared on the LSE Politics & Policy blog, Rod Hick looks at comparing multidimensional and income poverty measures.
Does measuring poverty multidimensionally make a difference in terms of who we identify as being poor? In recent years, a growing number of analysts have called for poverty measurement to go beyond a focus on income alone, to consider a wider range of deprivations a person may experience. The thinking behind such calls is typically that there are many ways that a person’s life can be impoverished, and that these need to be captured in poverty assessments.
Assessing poverty and deprivation multidimensionally proves to be more complex than just looking at household income, however. In quantitative analysis, we’re often limited by the information collected in major household surveys: you can’t count what isn’t measured. Then, the selection of which dimensions to include in the assessment often proves to be controversial, as the UK Government found when it mooted a suite of new child poverty measures that included ‘family stability’ amongst the dimensions for consideration. It can also be difficult to summarise performance on distinct dimensions – say, ill-health and housing deprivation, without appearing to add up ‘apples and oranges’.
In their 2011 book Poverty and Deprivation in Europe, Brian Nolan and Christopher T. Whelan argued that ‘“the need for a multidimensional measurement approach in identifying the poor/ excluded is an empirical matter, rather than something one can simply read off from the multidimensional nature of the concepts themselves” (p.19). Their point was that the belief that poverty has many dimensions does not necessarily mean that we must measure each of these when seeking to identify the poor or excluded. A multidimensional measure might identify the same people as being poor as an income poverty measure, set at 60% of the national median income, for example. Therefore, any limitations of income poverty measures must be demonstrated empirically, rather than assumed: a multidimensional analysis must demonstrate some distinctive empirical features in order to justify this more complex approach over-and-above a focus on income poverty alone.
In a recent study, I assessed the distinctiveness of multidimensional analysis of poverty and deprivation by comparing material poverty and multiple deprivation in Britain, drawing on data from the British Household Panel Survey.
Material poverty is measured by two indicators: relative income poverty (where a person’s income falls below 60% of the national median) and material deprivation (where people cannot afford one or more of a set of nine necessities, such as having two pairs of all-weather shoes, or being able to keep their home adequately warm). Multiple deprivation is measured across seven dimensions: general health (where respondents indicated their health was poor or very poor), mental health, housing deprivation, a lack of autonomy, life satisfaction, financial stress (finding it ‘quite’ or ‘very’ difficult to makes ends meet), and unemployment.
The analysis explores the relationship between each of these dimensions individually, as well in terms of the aggregate constructs of material poverty and multiple deprivation. In the aggregate analysis, the experience of relative income poverty or material deprivation is considered indicative of material poverty. The experience of any two or more of the seven dimensions listed above is considered indicative of multiple deprivation.
The analysis shows that the distinctiveness of multidimensional analysis depends crucially on whether we are interested in identifying vulnerable individuals or vulnerable groups and whether we focus on the individual dimensions within material poverty and multiple deprivation, or on the aggregate constructs. The nine measures identified substantially different individuals as being poor or deprived. This was observed irrespective of whether we focussed on the individual or aggregate measures.
We can see this in the table, which demonstrates this mismatch between material poverty and multiple deprivation, using the aggregate measures. Of the 26.2% of respondents experiencing material poverty, just one-third (or 8.3% of the total sample) also experienced multiple deprivation. These aggregate measures identify quite different individuals.
There is greater consistency between the measures when our focus is on identifying vulnerable groups, though the dimensions continue to provide distinctive results when the dimensions are analysed individually. Some groups, such as social and private rented sector tenants, or respondents in workless households, displayed consistently elevated risks of material poverty and multiple deprivation across each of the dimensions examined. Yet for others, such as single parents, respondents with low levels of educational attainment and families with children, the elevated risks they face in terms of the two dimensions of material poverty were not consistently observed when one looked across the dimensions of multiple deprivation.
On the other hand, when analysing the aggregate measures of material poverty and multiple deprivation, and seeking to identify vulnerable groups, there is a remarkably high level of consistency. As we can see in the graph below, most groups experiencing high rates of material poverty, such as single parents, social housing tenants, or workless households, also experienced elevated rates of multiple deprivation (indeed, the correlation of material poverty and multidimensional deprivation scores for 35 population sub-groups is 0.92). When our interest is in aggregate performance, the inclusion of multiple deprivation adds little to our understanding of which groups are most at risk.
The analysis shows that there are novel findings that flow from a multidimensional analysis that cannot be obtained from focusing on measures of material poverty alone. But, distinctiveness is not an all-or-nothing affair. The “value added” of multidimensional analysis depends on whether we analyse aggregate or disaggregated dimensions, and whether we seek to identify vulnerable individuals or vulnerable groups. These are different, but important, tasks for public policy.
Rod Hick is a Lecturer in Social Policy at Cardiff University. His primary research interests are the conceptualisation, measurement and analysis of poverty; the capability approach; and social security. This blog was originally published on the LSE British Politics and Policy website, and draws from a paper recently published in the Journal of Public Policy.