Disability Estimates

Overview

Many housing providers are aware that the details that they have in their data may massively underestimate the true level of disability within their resident population. This inference might be drawn from the House of Commons Library Research briefing entitled ‘UK disability statistics: Prevalence and life experiences’ (July 2022) which provides an overview of expected disability rates which may well be much higher than the recorded disability amongst a resident population. The briefing points to the most useful resources for measuring disability, namely benefits data, census data and the Family Resources Survey (FRS) from National Statistics and that is what we exploit in this dashboard.

The FRS is ‘considered the primary measure of disability prevalence in the UK. FRS disability estimates are based on respondents self-reporting “a physical or mental impairment” which has “a substantial and long-term adverse effect” on their ability to carry out day-to-day activities, in accordance with the Equality Act 2010 definition (p6, in House of Commons Report).

Detailed analysis can be used to provide robust (statistically valid) estimates of the likely level of disability within any resident population by using the FRS to generate estimates for a population that is as close as possible to that population. Note that we use the terms ‘difficulty’ and ‘disability’ interchangeably, reflecting their use in the different sources.

The dashboard consists of just two screens:

  • a simple overview of expected numbers of people with a disability in any population - just requiring the number of residents and primary region

  • an opportunity to explore underlying rates for adults and children and the impact of different age bands, housing tenure, region and gender on different types of disability (difficulty).

Numbers

The bar chart and table shows the estimated number of people with different types of disability based on the numbers that the user entered in the input boxes (in this example 10,000 adults and 5,000 children). Note that these are only estimates and not exact numbers. They are based on the results of surveys. The level of accuracy can be understood by exploring the underlying rates (click on the Explore Rates button) as these include confidence intervals.

The estimates are based on two of the six housing tenure types, namely Rented for Council and Rented for Housing Association. These have very similar rates and the survey documentation suggests that many residents are not aware of the distinction, hence we have grouped them to help ensure a good sample representation.

The bar or table row that refers to ‘Any one or more’ means those who have at least one difficulty. Many respondents to the survey would have had two or three difficulties, so the number of people with difficulties is not simply a sum of all the different types of difficulties.

There is an option to limit the numbers to only those who experience a significant impact. The survey asked the question: ‘Does the condition limit day to day activities?’. By default all levels are selected, but the user can select one or more of these categories to focus attention on those with more more need. This will lower the expected numbers.

Click on the age band labels at the top of the table, the chart will then display that age band.

A second option allows the user to input the adult numbers by ten year age bands. This can improve the accuracy, for example if the residents are predominantly older people. If these bands are not used, the dashboard will take a default population distribution based on the region.

Rates

The aim of the page on rates is to enable the more curious user to understand the variability of difficulty rates based on key dimensions, namely:

  • Age Bands

  • Region

  • Tenure Type

  • Gender

This chart is an excellent example that demonstrates the massive impact that housing tenure type has on difficulty rates.

Note the lower and upper limits which are important indicators of accuracy. As the survey is just a sample of the adult population, we cannot be sure that the respondents were fully representative of the whole population. We constructed 95% confidence intervals which means that we can be 95% confident that the actual population of the UK has rates in this range (or to express it the other way round, there is only a 1 in 20 chance that the ‘true’ rate lies outside this range).

In exploring these rates, note how the confidence bands get very large if the effective sample becomes too small to draw any meaningful inferences, generally if the count goes below about 20 individuals. By selecting categories in the filter pane (age bands, region, tenure type, gender) it is easy to drop the sample down to a small number. The actual numbers in the filtered survey are shown in the big number on the top right of the chart. We also include the gross number of adults that the filtered sample represents in the total population. In the example below, if we select the Age band - 65+, Region - London, 2 Tenure Types (Rented from Housing Association and Rented from Council), Gender - Male, we have just 185 respondents. For a difficulty such as Mental Health, this results in just 4 respondents in ‘Rented from Housing Association’, so the confidence interval goes from 2% all the way to 16%. Mouse over each column to show how the rates are calculated.

Click on the options (Tenure Type, Age Band, Gender) to see how the survey data (whether filtered or not) on the ten types of difficulty breaks down by each of these categories.

A similar page is available for the Children, with their different age bands. Note that now there are no population estimates as there are no grossing values in the source survey for children. In other words the rates for adults are calculated from the grossed counts but, for the children from the survey counts.

The mouseover in both the adult and child screens shows the numerator and the denominator. If we apply a filter (such as the region) the numerator and denominator are filtered accordingly. However we can also filter the numerator only, by filtering the count of people with a difficulty by the impact. In other words we can limit the condition to those whose day to day activities are reduced a lot. This will lower the rates.

Department for Work and Pensions. (2023). Family Resources Survey, 2021-2022. [data collection]. UK Data Service. SN: 9073, DOI: http://doi.org/10.5255/UKDA-SN-9073-1