Book Chapter Inequality and Poverty Re-Examined Ch. 8
Summarizing multiple deprivation indicators
It is widely agreed nowadays that being poor does not simply mean not having enough money. It means, more generally, a lack of access to resources enabling a minimum style of living and participation in the society within which one belongs - as in the definition of poverty adopted by the European Union, for example. In short, poverty is not only about low income, but also about deprivation.
These are not simply academic concerns. Assessments of deprivation are fundamental parts of national anti-poverty strategies in several countries. Summary indices of deprivation are used in combination with measures of low income to produce pictures of 'consistent poverty' in the National Action Plan Against Poverty and Social Exclusion in Ireland. In the UK, progress towards the eradication of child poverty is to be monitored not only using income poverty measures but also with measures of 'material deprivation'. Deprivation indicators are included in the main EU surveys for social monitoring, i.e. the European Community Household Panel and the EU-SILC surveys, and are part of a wider portfolio of social indicators being developed at a European level.
This paper examines some methodological issues concerning the construction of a deprivation scale from multiple deprivation indicators, issues that have received little attention in the deprivation literature. We draw on the literature on item response modelling from psychometrics and educational testing as it has a long history of addressing similar measurement issues. Deprivation indicators are like test scores (i.e. whether an answer to a particular test question is right or wrong), and summarising deprivation indicators with a deprivation scale is like summarising test scores with a scale of academic ability. Our particular interest is in assessing the ubiquitous practice of constructing a deprivation scale as a raw (or weighted) sum of a relatively small set of dichotomous indicators.
We argue that the theoretical foundations of these 'sum-score' scales are relatively weak and that the item response modelling approach provides a more promising way to summarize multiple deprivation indicators. An application based on British Household Panel Survey data is used to illustrate the arguments.
We focus on 'basic life style' deprivation, summarized using seven binary indicator variables. The first six variables summarize responses to questions put to the household reference person asking whether he or she would like to be able to PHRASE but must do without PHRASE because they cannot afford it (an 'enforced lack'), where PHRASE refers to:
Keep your home adequately warm
Eat meat, chicken, fish every second day
Buy new, rather than second hand, clothes
Have friends or family for a drink or meal at least once a month
Replace worn out furniture
Pay for a week's annual holiday away from home
Each variable was scored one if there was an enforced lack of the relevant item or activity and zero otherwise; the percentage in parentheses is the fraction of the sample with an enforced lack. The seventh binary indicator variable summarized difficulties in meeting housing costs: i.e. whether the responding household had any difficulties paying for their accommodation in the last twelve months. Those reporting payment problems scored one on this variable; otherwise it was zero.
As it happens, both the item response modelling and the sum-score approaches provide very similar pictures of the patterns of basic lifestyle deprivation and their determinants, and so our results might be construed as providing an empirical rationale for the sum-score approach. We address this issue in the final sections of the paper, where we combine further discussion of the relative merits of sum-score and item response modelling approaches with suggestions of ways in which the latter approach could be developed further.
Volume and page numbers
166-184 , 166 -184
by Stephen P. Jenkins and John Mickelwright (eds.)