Using Census data to derive a new area-based measure of deprivation - Section 1: Introduction

Updated 19 March 2024

Section 1: Introduction

The fundamental purpose of this paper is to present a new UK-wide area-based measure of deprivation derived using Census 2011 called the Socioeconomic Index for Small Areas (SEISA). Across all nations of the UK, reducing regional inequality continues to be a key policy objective. To meet this goal, resource needs to be allocated to those parts of the country where they are most required. However, this opens up the question of how one defines and identifies the most deprived localities. We therefore begin with a discussion of how previous literature has interpreted the term deprivation, before looking briefly at what measures currently exist, including their limitations.

Townsend (1987) specified this concept as one in which an individual or household has clear disadvantage relative to other members of their local community or wider society. A similar relative construct was put forward by Gordon (1995) where those in deprived conditions were seen as having a level of resource that prevented them from fully participating in well-established activities and norms in public life. More recently in policy circles, the Scottish Government (2018) have set up the Fairer Scotland Duty, which requires public bodies to reflect on how they can reduce the extent of inequalities that result from socioeconomic disadvantage. As in the academic literature, a relative definition has been put forward, where those in deprived circumstances are seen to be living in inferior social and economic conditions compared with others in the same society.

With some general alignment over what is meant by deprivation, the next issue for consideration is how it can be measured. In the UK, material deprivation (i.e. whether an individual/household possesses certain material things) is currently captured through the Households Below Average Income dataset, which is itself derived using the Family Resources Survey. However, this only covers a sample of households across the UK, meaning it cannot be used to understand the levels of deprivation across all areas of the country. Additionally, exploring matters such as social deprivation is also not feasible, due to an absence of UK-wide information on the affordability of social activities. However, as an alternative, low income could be used as a way of identifying those experiencing deprivation. As Townsend (1987) noted, those facing one or more forms of deprivation have a very high probability of having little financial resource. However, income data too is not typically available in the public domain at either individual or small area level. This has left researchers having to use national collections such as the Census to formulate indices that are likely to be highly associated with deprivation.

Townsend (1987) highlighted the potential for four variables (available in the Census) to be utilised in the formation of an index relating to material deprivation. These were unemployment, overcrowding, as well as car and home ownership. Yousaf and Bonsall (2017) have constructed the Townsend Deprivation Index using the 2011 Census at various geographic levels, including output area. The index designed by Carstairs and Morris (1989) also utilised similar Census variables, though social class was included as opposed to a variable on home ownership. Both of these measures though have been subject to criticism. For example, the inclusion of a variable on vehicle ownership when constructing an index has been questioned by Allik et al. (2016) on grounds that purchasing a vehicle may be necessary in rural areas with limited public transport, even among poorer households. The reduced validity of indicators of socioeconomic deprivation that are based partly on vehicle ownership among rural localities was also illustrated by Christie and Fone (2003).

These are not the only two area-based measures of deprivation to have been developed. The most common index drawn upon by government departments in the four nations today in their decision-making processes is based on the Indices of Deprivation, with the work of Townsend (1987) providing the conceptual framework for the generation of these indicators. Originally created to support policy design - such as the allocation of resources for the Labour government’s neighbourhood renewal agenda in the early 2000s - the Indices of Deprivation initially covered domains such as education, employment and income. These were then weighted to generate a final index. Since then, they have continued to evolve every few years.

Currently, each UK nation creates their own final index and the different methodological approaches employed mean that this is not a UK-wide deprivation metric. For example, in Wales, eight domains are drawn upon to generate the composite measure, whereas Northern Ireland, Scotland and England utilise seven indicators to form their index. Weights assigned to similar domains also vary across countries. In Wales, community safety contributed 5% to the 2019 index, while in England, crime was given a weight of 9.3% when creating the final index. In their review of statistics in the post-16 education and skill sector within England, the Office for Statistics Regulation (OSR, 2019) noted the lack of a UK-wide deprivation metric was inhibiting the potential to publish country-level analysis, as well as assessing the progress being made on improving social mobility within society.

With regards to the size of the areas utilised in the Indices of Deprivation, lower layer super output areas (LSOAs) are presently used in England and Wales, which average approximately 1,500 inhabitants. Data zones are the geography level employed in Scotland, which have populations of 500 to 1,000, while in Northern Ireland, super output areas (SOAs) will generally consist of around 2,000 people. Areas are partitioned into deciles or quintiles, through which those living in the most deprived parts of a country are identified. It should be noted though that another limitation often raised about the Indices of Deprivation is that the size of the areas used can lead to it being difficult to identify pockets of deprivation within less deprived areas. Additionally, a common critique of the final index developed in different countries is that they fail to adequately capture deprivation in more rural aspects of a nation, where it may not be as geographically concentrated (as is often the case in urban areas). Indeed, the Commission on Widening Access (2016) report acknowledged this weakness, as did the Welsh Government (2015) paper on evaluating deprivation in rural areas.

Consequently, there remains an opportunity to create a UK-wide indicator that can suitably proxy for deprivation and that addresses some of the drawbacks of existing measures. We therefore utilise the 2011 Census to create SEISA - a new index at output area level based on the educational qualifications and occupations of residents. Output areas are the base unit for Census data and therefore represent the smallest geographical level at which aggregated statistics are released into the public domain. They are more commonly referred to as ‘small areas’ in Northern Ireland. They generally contain less than 500 individuals, hence using areas of this size is designed to mitigate the limitation of the Indices of Deprivation of not being able to pick up pockets of deprivation in otherwise less deprived neighbourhoods. The methodology we employ to develop our measure of deprivation shares some similarities to the approach adopted by the Australian Bureau of Statistics (2016), who draw upon their Census results to create the Index of Education and Occupation (IEO). The geographical unit they utilise to generate this index is the Statistical Area Level 1, which typically contains between 200 and 800 people. This composite measure is frequently used in Australia for decisions relating to funding, as well as for research/statistical purposes. The relation our measure has to low income and other features associated with deprivation, such as poor health, is also illustrated. Furthermore, we undertake a detailed comparison of how the variable we have generated differs to the Indices of Deprivation, including its ability to capture more rural localities.

The rest of this paper proceeds as follows. Section 2 outlines the data sources we use, while section 3 discusses our methodology and presents some summary statistics on our measure. We then go on to look at whether our measure is likely to be correlated with factors associated with deprivation and also conduct a sensitivity analysis to examine whether it could be improved by incorporating housing tenure. This is followed by an assessment of the variable, alongside the Indices of Deprivation. The study closes with concluding remarks.

Next: Section 2: Data