There’s a bunch of frequently used area-based measures of deprivation in higher education data.
Let’s name them.
To start with, the various national “indexes of multiple deprivation” (IMD, WIMD, SIMD,NIIMD) combine some 37 measurements across seven domains. As far as there is such a thing these are the government standard measures of deprivation – and we are increasingly seeing them in higher education data.
Then you have higher education specific measures – UCAS’ MEMs (which combines a bunch of information about educational experience, personal characteristics, and even IMD itself) is probably one of the more sophisticated compound measures, but you don’t often get to see it in use. It works at both an area based and individual level.
POLAR (and latterly TUNDRA) are straight measures of likelihood of higher education participation – these are not, strictly speaking, measures of deprivation but often work as decent proxies. People from deprived backgrounds do tend not to go to university, but there’s been an increasing amount of concern about the way these two measures are used in targeting access initiatives..
So OfS’ recent work on deprivation has tended to present individual measures in disaggregated form, and to look at individuals rather than groups – though we do see IMDs turn up a lot, the dominant approach would be something like what we see in the Association Between Characteristics (ABCs) work. Though there are ABCs quintiles, we are encouraged to look at the detail: parental higher education, national measures like IDACI – which covers childhood deprivation and NS-SEC – based on the job your parents do, and free school meal eligibility.
Free to choose
Your choice of definition is going to be dictated by what you are trying to do. There’s a body of thought that suggests you should use data that is most closely aligned to the question you are asking – so POLAR or TUNDRA would be the right way to examine disparities in access. Others contend that intersections of multiple measures can show evidence of pockets of deprivation. Confusingly, both of these are true – there are individualised indicators that point to a set of life experiences that are unlikely to lead to higher study, but the fact that we are interested in higher study as an endpoint means we can quite closely target local areas where this doesn’t happen without having to know everyone’s personal characteristics.
The middle ground would be to look at area-based measures of deprivation rather than higher education participation – which is what OfS is doing with IDACI and IMD, and what HESA are doing here.
One way out of this confusion is to do what HESA has done this morning – look for instances where one measure accurately predicts a bunch of others – a proxy measure, in other words. Tej Nathwani and Siobhan Donnelly have found that the classic Index of Multiple Deprivation (IMD) can be closely approximated by a much simpler composite variable based on the qualifications of residents in an area and their occupation (NS-SEC). Both these are taken from the 2011 Census: specifically the proportion of people in an output area in NS-SEC groups 3-8 (so, any occupation other than a managerial or professional one) and the proportion aged 16 and over with their highest qualification below Level 4 (so sub-higher education).
Podcast fans may be aware that I sometimes run a podcast segment called “yes, but does it correlate?” – similarly HESA have found a very clear statistical relationship (Pearson’s coefficient of 0.91) here. This make sense – you tend to need a high level of skill to work in a skilled job.
But not only do these two work nicely together, if you take an average of both you can be quite accurate in identifying areas with low income and all the other stuff in the IMD., and because we are calculating from UK wide variables we don’t face the issue that we do in having four incompatible IMDs – we can compare areas across the whole of the UK.
The big benefit is that, in every case, HESA’s new measure is better at identifying rural deprivation than the IMDs. It’s not perfect – we’re dealing with decade old data, at least until we get a full release from Census 2021 (and Census 2022 in Scotland) – but for anyone planning outreach work it is a very promising starting point. HESA will be publishing more information, and more data, in the weeks to come.