By Ian Smith
There has been a lot of interest in measuring disadvantage over the past 20 years in the UK even if this has not always been matched by government responses. The fifth iteration of the English IMD is to be reviewed over the next 12 months. Clearly disadvantage is a complex thing and can be represented in many different ways. As a geographer (or someone who periodically claims to be a geographer hiding in an Economics Department) I am particularly interested in area-based assessments of disadvantage. I know such measures are problematic but what indicators are not? I recently have had the opportunity with colleagues to review how the English Indicator of Multiple Deprivation works on behalf of Power to Change (see https://www.powertochange.org.uk/) and this is a short blog that captures some of the thinking that came out of that work (any errors or misinterpretations are all my/our fault and not necessarily shared by anyone at Power to Change).
So, the English IMD is a second-generation indicator of area-based deprivation that represents 7 ‘dimensions’ (or 10 sub-dimensions if you like) of disadvantage from worklessness to housing affordability, from health (mental and physical) to distance from your nearest post office. It is ‘second generation’ because it is not solely dependent on small area census data (as ‘first generation’ indices are/were) but is based on a range of small area administrative and census data from different sources within English government.
I am a fan. It is lovely. My colleagues in other European countries are jealous of it (the basic model is oft copied) – both because of its breadth of content but also because of our lovely regular statistically ordered lower super output areas (LSOAs) that sometimes get conflated for neighbourhoods. However, an indicator is a conceptual model of a real concept. As George Box pointed out – all models are wrong, but some [of the better ones] are useful. We and Power to Change were interested in posing the question of how useful is the IMD to Power to Change?
In particular, we were interested in how the IMD is used within a particular organisational context (Power to Change). We set up a set of dimensions to help us think about how an indicator (a statistical instrument ‘designed’ to perform a task) is constructed and deployed. We asked people in Power to Change how they used the IMD and what was their assessment of the strengths and weaknesses for what they needed to do: investing in community businesses that alleviate disadvantage in England. What struck us in these conversations was that the IMD was only being used in its top-line indicator format – what was being missed was the opportunity to use the IMD as an indicator system that can be moulded to the specific objectives of an organization.
We explored how to use the IMD as a system of indicators to shine a light on a specific objective: investing in community businesses. We compared spatial targeting at LSOA level for the top-line IMD indicator (the full 7 dimensional one) with the spatial targeting from a bespoke indictor bringing together the health and disability, education and qualifications and the geographic access to services dimensions. Power to Change has hypothesised that community businesses some of which provide local services may impact on employability (skills) and on the health of residents in the communities that community business serve. So, we constructed a focussed indicator from components of the topline IMD that focused only on geographic access to services, education and health (for details see Smith et al 2018). We compared how the focussed IMD indicator would spatially target the attention of Power to Change in comparison to the top-line IMD indicator with a particular focus on the city-region of Liverpool and the County of Suffolk as examples of areas of interest for Power to Change. We then mapped out the differences (using data and shapefiles obtained under a public licence) showing firstly the map of the top-line IMD indicator, secondly showing our ‘new’ indicator focusing on Power to Change’s priorities and thirdly what difference it makes in targeting. These maps are shown in Figures 1 (for Liverpool) and Figure 2 (for Suffolk). We have used the somewhat arbitrary threshold of 30% to indicate disadvantage (the most disadvantaged areas to be targeted) and compared the indicators.
The left-hand side map in both Figure 1 and Figure 2 shows neighbourhoods marked relative to the top-line IMD indicator where the deepest green areas are the most disadvantaged. In the middle map the same rule applies. The right-hand map in these Figures shows what difference it makes for these areas. In this right-hand map, the red areas are those that are marked as the most disadvantaged 30% under both indicators. The blue areas are ‘advantaged’ under both measures. However, the orange areas are marked as disadvantaged under the ‘better places’ indicator but not under the top-line IMD.
Given the greater importance given to access to services (albeit direct distance accessibility based on 2012 data), it is not surprising that Suffolk LSOAs become more disadvantaged under this measure. Thus, nearly half of Suffolk becomes ‘disadvantaged’ on this measure (30% most disadvantaged in England on this measure) than under the top-line IMD (more of Suffolk’s third map is coloured orange). Perhaps it is of greater surprise that the prioritisation of Liverpool changes little under the new formulation. Most of Liverpool’s neighbourhoods remain identified as ‘disadvantaged’ (marked as red in the third map along).
This is however, just a schema for moving resources around. It is an inevitable result of re-calculating the target IMD measure that some areas gain whilst others lose out (where resources are fixed). However, if areas in Suffolk gain whilst neighbourhoods in Liverpool do not lose out, then how would such a change modify the geography of disadvantage [under this measure] across England? Using the 30% figure as the threshold of disadvantage just under half a million fewer people would be designated as living in a ‘disadvantaged’ area. We did some cluster analysis of the ranking on the top-line IMD indicator and our suggested Power to Change indicator considering both how LSOAs clustered together (using forms of hot spot analysis) to capture how patterns of disadvantage form broad regions and secondly, we looked at the identification of outlier neighbourhoods (using the analysis of Anselin Local Moran’s I) to capture differences within these wider clusters.
On Figures 3 and 4 the LSOAs that are marked as red are ones than appear as advantaged (close to other advantaged areas). In these Figures we have a left-hand map that shows the clustering of indicator ranking in relation to Suffolk. The middle map shows the Getis-Ord clustering for England as a whole whilst the right-hand map shows the Local Moran’s I maps which show where areas are located as outliers in wider regions. Where there is red there is advantage and where there is blue there is disadvantage (from an area-based perspective). Yellow areas are mixed (any area’s ranking is not easily predicted from the ranking of its neighbours). It is also worth noting that the red and the blue areas are not necessarily all of the most disadvantaged areas – just areas that are close to others that are similarly ranked (whether high or low).
It is not surprising to see clusters or disadvantaged (blue) areas in England’s northern metropolitan areas, in the West Midland and in the extreme South West in Figure 3 that maps out the top-line IMD indicator. It is also not surprising to see the East and
North of London marked as deep blue although it is worth noting that the former Kent Coalfield areas remain marked as disadvantaged in blue. So, it is England to the south of the Wash to Severn axis as well as North Yorkshire that are marked as ‘advantaged’ regions under the top-line IMD indicator. The Anselin outlier mapping (right hand map) in Figure 3 points out the presence of disadvantaged LSOAs in advantaged clusters and of the presence of advantaged LSOAs in disadvantaged clusters.
Moving to the Power to Change indicator in Figure 4 we see a change in the geography that might be targeted (in this case by investment in community businesses). More rural areas in the East and South West of England become identified as ‘disadvantaged’. Areas in the East and North of London no longer become identified as disadvantaged in terms of the clustering on this measure’s ranking. There is a different dynamic – to be disadvantaged area in London is to be surrounded by advantaged areas. The East of England (including Suffolk) becomes identified with the cluster of disadvantage although there are clearly still advantaged area outliers in the sea of blue disadvantaged areas. Although there are disadvantaged areas in the advantaged region of London. It has to be stressed that this applies only to forms of disadvantage that flow from combinations of problematic educational, health and accessibility outcomes. There would be a case for an organisation like Power to Change to use a form of IMD that relates specifically to their core mission as a spatial guide to targeting rather than just using the top-line IMD indicator.
The aim of the exercise is not to rubbish the general top-line IMD. I am still a fan – it is still offers useful insight into the patterns of generalised area-based disadvantage across England. The English IMD is still useful to Power to Change in a general sense. However, the aim of this has been to draw to attention the fact that deploying the indicator system in the light of what is trying to be achieved makes better use of the IMD system. The East and North of London is clearly a region with many disadvantaged areas but if the aim of the exercise is to invest in community businesses that improve access to services, health and educational outcomes, there might be better areas on which to focus this specific form of investment. Whatever form of analysis we come up with to capture disadvantage there is always a set of political choices about how to share out public spending. However, the English IMD is more than just the top-line indicator and the top-line IMD was never intended to be the only way in which area-based disadvantaged was represented.
Although in this delicate dance of spatial targeting, the real answer is to invest more in welfare services. Perhaps that is one normative step too far?
If you want to read more about our work with Power to Change, please download the report we wrote for them (available from September).
Smith, I, Green, E, Whittard, D. and Ritchie, F. (2018) Re-thinking the indices of multiple deprivation (for England): a review and exploration of alternative/complementary area-based indicator systems. Final Report. Bristol Centre for Economics and Finance (BCEF) in the Bristol Business School at the University of the West of England (UWE).