Using Census data to derive a new area-based measure of deprivation - Summary

Introduction

In October 2021, HESA released a paper detailing a new UK-wide area-based measure of deprivation based on Census 2011 data. We requested feedback from users on the potential value this could bring, as well as any further exploratory work they would like to see from us.

The sector has been positive about the usefulness of the measure we have created, with many of the comments asking for HESA to conduct supplementary research to support its development. A summary of the requests is outlined below.

• More information on why education and occupation have been used to generate the measure and how we reached the conclusion not to include aspects such as housing tenure and car ownership.
• Where possible, additional analysis that illustrates the extent to which our measure is correlated with income (ideally equivalised and net of housing costs).
• Further comparisons with existing area-based measures, including the individual components of the Indices of Deprivation and an Income Deprivation Affecting Children measure (where available).
• Those working in the English higher education sector also wished for us to undertake an assessment of our measure against TUNDRA (tracking under-representation by area).

Today, we publish an updated paper that seeks to address the feedback submitted. Below, we provide a brief question and answer to summarise the key points. We thank those who reviewed and provided helpful comments on earlier editions of this work.

Who should read this paper?

Though the remit of HESA is to publish statistics relating to higher education, we appreciate that this deprivation measure we have created could be valuable beyond our own field. This work will therefore be of interest to those who carry out research/analysis relating to deprivation, as well as those who require such information for their decision-making processes. For example, an indicative list of who our study may prove useful to include those within;

• Other parts of the education sector
• National Government
• Local Government
• Industries such as healthcare
• Charities/researchers/other organisations specialising in issues relating to poverty, deprivation and social mobility

Why do you use the Census to develop an area-based measure of deprivation?

Those experiencing deprivation will have a level of resource that prevents them from fully participating in well-established norms and activities of public life. However, to the best of our knowledge, no data exists on the affordability of such activities. While it is highly likely that those in severe deprivation will have low levels of income, there is also an absence of accessible income information at individual and small area level.

It is for this reason that researchers have to rely on collections such as the Census to develop suitable proxies for deprivation.

Does the analysis in this working paper again link Census information to HESA records?

We no longer restrict our work to higher education students and therefore no HESA data is drawn upon in this piece. Indeed, no individual-level information is utilised in the research.

Rather, output area is the smallest domain on which our analysis is based and we ensure to cover all 232,296 output areas across the UK. This may not necessarily be the case when the sample is limited to higher education students, as there may be some output areas where no individuals have chosen to pursue a degree.

Data at output area level from the Census is linked to higher level geography information, which is then used to ingest the Indices of Deprivation, as well as other useful data (such as urban-rural classifications). See section 2 of the paper for further details on our data sources.

What are the advantages of not linking to HESA records?

Carrying out an investigation that extends across the entire of the UK can assist with further validating the value of our measure.

How did you determine that education and occupation data from the Census should be used to develop an area-based measure of disadvantage?

A paper released by HM Government (2014) illustrates the key role that low qualifications and poor employment outcomes play in individuals/households experiencing long periods of low income. These two factors are also crucial when trying to understand why poor children may remain poor during adulthood.

The validity of using car ownership in developing a measure of disadvantage has been criticised on the grounds that those in rural localities may find it a necessity to purchase a vehicle, even if they are living in deprived circumstances.

The HM Government (2014) study also notes that it is low income that is more likely to result in someone living in poor quality housing rather than the other way around. Given recent decades have seen soaring rental costs and stagnating wages, we did examine whether housing tenure could improve our measure, but the theoretical and empirical evidence does not appear to support this. Section 5 of our paper provides more detail.

To what extent is the measure correlated with low income?

For England and Wales, we have net equivalised weekly household income estimates (after taking into account housing costs) at the middle layer super output area (MSOA) level. We use this to illustrate how those areas that are defined as more deprived under our measure do appear to also contain households with lower levels of income.

The Census does contain UK-wide information on aspects such as self-reported health and household composition. Research suggests that poorer self-reported health and being in a lone parent family is correlated with low income. We demonstrate the most deprived output areas according to our measure also have a greater proportion of lone parent households/residents who report poorer health.

Section 4 of our paper supplies further details on these matters.

How does your measure compare to the Indices of Deprivation?

Firstly, the index formed from the Indices of Deprivation is a nation-specific measure and is therefore not UK-wide. That is, statistics based on the index cannot be compared across nations. The reason behind this is that each country adopts a different methodology to generating its index. For example, the Welsh Index of Multiple Deprivation is created using eight domains, whereas seven are used in the other UK countries. Furthermore, how each domain is weighted in the final index also varies by country. For instance, community safety contributes 5% to the final index in Wales, whereas the corresponding figure for the crime domain in England is 9.3%.

A key difference with our measure is that using the Census has enabled us to develop a UK-wide measure of deprivation. Questions included in the Census are harmonised as far as possible across the nations and we use the same geographic domain across all four nations to create our measure (i.e. the output area). In Northern Ireland, they were referred to as ‘small areas’ in the 2011 Census, though these were kept identical to the 2001 output areas where this was feasible to assist with comparability over time.

Another acknowledged limitation of the index of multiple deprivation across the nations is that it is less effective in capturing deprivation in rural areas.

In comparing (the bottom quintile of) our measure to (the corresponding quintile of) the composite index in all four nations, we find that in each country, our measure captures a larger proportion of deprived areas that are classified as rural locations.

We do not generally find that our results change if we focus on the individual components of the Indices of Deprivation, such as income. The only exception to this is in Northern Ireland, where the income component of the index, as well as Income Deprivation Affecting Children do buck the general pattern.

Aside from this, we can summarise the added value our measure brings in each nation as follows;

England: The bottom quintile of our measure picks up a greater proportion of medium/large towns in northern and central parts of the country where earnings and attainment at Key Stage 4 are often below the national average.

Wales: One of the key differences between (the bottom quintile of) our measure and the Welsh Index of Multiple Deprivation is the ability of our variable to catch deprivation in two local authorities of south Wales (Rhondda and Caerphilly).

Scotland: All council areas in Scotland emerge in the bottom quintile of our measure, which is not the case with SIMD. Na h-Eileanan an Iar, as well as Shetland and the Orkney Islands have no areas that appear in the lowest quintile of SIMD.

Northern Ireland: The benefit of our measure relative to the Northern Ireland Multiple Deprivation Measure (based on our assessment of the lowest quintile) appears to lie in its capacity to pick up parts of the north and/or east of the country.

How does your measure compare to TUNDRA?

We do not discuss TUNDRA in the paper itself, given that it is not designed to be a measure of socioeconomic disadvantage. Rather, it focuses on the likelihood of entering higher education. More information on TUNDRA can be found on the Office for Students website. There is also no similar measure available in the other nations of the UK.

However, given the interest shown by users in the English higher education sector in us undertaking a comparison between our measure and TUNDRA, we do so here.  

As we state in section 6 of the paper, the mean net equivalised weekly household income estimates (after housing costs) in England for the bottom quintiles of our measure and the English Index of Multiple Deprivation is around £360. The equivalent figure for the lowest quintile of TUNDRA is £393.02. TUNDRA therefore does appear to be less correlated with low income.

In a similar vein to POLAR (participation of local areas), the lowest quintile of TUNDRA captures a very low proportion of students from London in the bottom quintile. Overall, the lowest quintile of our measure captures a greater proportion of output areas in central and northern parts of England, while the bottom quintile of TUNDRA picks up a larger percentage of the south of the country.

How do you anticipate your measure being used by the higher education sector?

Many higher education providers run outreach activities in which prospective students must meet one or more eligibility criteria to be able to participate in the programme. One of these conditions will often be that the student should be living in an area of socioeconomic disadvantage.

However, as we discussed earlier, existing area-based measures have their limitations (e.g. not effectively capturing rural locations), meaning there will be individuals living in deprived areas that do not meet the eligibility criteria stated to take part.

We hope that our measure can soon be added to the list of eligibility criteria used by providers to ensure that those living in deprived areas that do not emerge in the bottom quintile of the index of multiple deprivation do not miss out on such schemes.

This will assist with the policy objective of promoting equal opportunity for all in the higher education sector.

For some providers, the measure we have developed may help in determining some of the localities that may benefit from higher education outreach and hence where they may wish to focus their resources.

With raising aspiration and attainment often seen as important features of widening participation activity, we have shown in earlier research that some of the areas that are picked up in the bottom quintile of our measure, but not in the equivalent quintile of the index of multiple deprivation for a nation, tend to have lower than average attainment levels.

How do HESA plan on using the measure in their work?

The Office for Statistics Regulation have previously noted that the lack of a UK-wide deprivation metric inhibits comparable country-level analysis to be published and to understand the wider trends in social mobility. Indeed, a similar point was raised on page 20 of the State of the Nation 2022 report published by the Social Mobility Commission (2022).

It is hoped that the continued development of this measure will enable us to enhance our official statistics outputs and address this issue raised.

What are the next steps?

This can be summarised as follows;

• Explore how we can make the measure accessible to the sector, so that it can be utilised as part of widening participation activity.
• Examine the possibility to obtain individual-level data to further validate the correlation between our measure and low income, as well as explore the potential to create a ‘mixed metric’ that combines individual and area-level information into a single composite variable.
• Update the work we have done using Census 2021 (this is likely to be a few years away, particularly given Scotland conducted their Census in Spring 2022).

Comments/feedback on this updated paper are welcome and can be sent to [email protected].

Next: Main paper - Abstract