There are a number of different methods that can be used to aggregate the newly-published 2025 English Indices of Deprivation measures to wards (and other custom geographies).
Those using Local Insight to make it easy to explore the Index of Multiple Deprivation (IMD) ranks for their wards may see differences when comparing outputs from the platform to other sources of the same data.
These changes come as a result of the aggregation and apportioning used within Local Insight. Our methods follow the same principles behind the national Indices of Deprivation approach, but include a few steps that mean that the figures do not always match a simple average of a ward’s Lower-Layer Super Output Areas (LSOAs).
This blog explains the full method that Local Insight uses, from the published LSOA scores through to the final ward values, and then explores how Local Insight represents Local Authority-level deprivation data.
LSOAs do not fit neatly into wards. Below is the Bathavon North Ward in red, with the closest-matching LSOA (Bath and North East Somerset 010C) in black.
In many places, an LSOA overlaps two or more wards. If we were to assign the whole LSOA to whichever ward contains most of its postcodes, it would build a mismatch into the results. That creates problems when comparing wards or identifying the most and least deprived areas.
There are different ways to avoid this, and Local Insight’s approach starts by first apportioning the LSOA IMD scores down to Output Areas (OAs).
OAs are the smallest Census geography, and can always nest inside LSOAs – meaning that they can provide a better building block for constructing and aggregating data to larger geographies, such as wards.
We begin with the official LSOA-level IMD scores from the English Indices of Deprivation 2025 (IoD 2025). At this point, all the data sits at LSOA level.
Each LSOA contains several OAs of different population sizes. The OAs that fit within an LSOA are called its ‘child areas’ – and these are given the same value as their ‘parent area’.
We do not weight the OA population at this step, because the parent value is already a population-weighted average.
For example, for an LSOA with an IMD rank of 1,000, all of the child OA’s within that LSOA would also receive a value of 1,000.
While they’re more accurate than LSOAs, OAs still do not always line up cleanly with ward boundaries, so we use the Office for National Statistics (ONS) best-fit approach based on population-weighted OA centroids.
The population-weighted centroid is the point within the OA that represents the average location of a population within an Output Area (OA), with a higher weighting given to more populated areas. For example, if there are several large apartment buildings within the OA, the centroid will be closer to them, rather than an equal-sized area of single-family homes.
What that means:
All population centroids and official lookups come directly from the ONS open geography portal.
This method ensures that each OA is allocated to the ward it best aligns with in population terms, rather than through postcode count alone.
Once every OA has an apportioned value, a census population and a ward assignment, we build up the ward rank using a population-weighted average:
Ward Value = ∑ (OA Population × OA Value) / ∑ (OA Population)
Deprivation measures reflect the experience of people, not land area. An OA with more residents should have a proportionally greater influence when we rebuild the score for a larger geography.
Since we’re aiming to produce one correct value for the parent area, and each child area may have different populations, we use population weights to ensure that each child area contributes fairly to the parent rate.
You may have noticed that the IoD figures for Local Authorities (LA) in Local Insight differ from the figures published by the Ministry of Housing, Communities and Local Government (MHCLG).
In their tables, MHCLG uses higher ranks to represent higher deprivation for LA’s, whereas at LSOA level lower ranks indicate higher deprivation. This is because the LA statistics are summary measures (such as average ranks) and MHCLG presents them in a “higher = more deprived” format for ease of interpretation.
To keep things clear and consistent for Local Insight users, we have reversed these LA ranks across the site so that, across all area types, a lower rank always represents higher deprivation – matching the familiar LSOA-level convention.
For example, if a Local Authority had a published average LSOA rank of 10,000, we would assume this area had higher deprivation relative to other LA’s using the known LSOA convention of “lower = more deprived”. However, this is not the case, and this area would actually have lower levels of deprivation according to MHCLG’s presentation. To make this easier to interpret, we minus this figure from the total number of LSOA’s in England + 1 (33,756) to create a new rank of 23,756 for this LA.
In other words, within Local Insight the Local Authority rank follows existing IoD rank logic, with a rank closer to 1 always representing a more deprived area.
Most indicators within Local Insight are downloaded from the source as counts – here, we call them standard indicators.
When aggregating these standard indicators, it’s as simple as adding up a number of child areas to find the value for the parent area. When apportioning these standard indicators, the values are apportioned down to child areas based on their population.
For example, when apportioning a count value from LSOA to OA, we follow this calculation: LSOA VALUE * (OA POPULATION / LSOA POPULATION).
After we aggregate these counts up to higher geographies and apportion them down to lower geographies, we create rates/percentages for each area using a suitable denominator (such as population or the number of households).
We use the term ‘non-standard indicators’ here, for any datasets where the rate or percentage is the only value we are able to get from the source. These are indicators such as the Indices of Deprivation, Average Life Expectancy and Age-Standardised Mortality Rates.
To aggregate these datasets up to higher geographies, we use the population of each child area to give the child value a weight, before dividing the sum of these by their total population to find the total parent area value.
When apportioning, all child areas have the same value as the parent area.
This ensures that all higher-level aggregates produced by Local Insight follow consistent rules and do not depend on boundary alignment.
As of November 2025, Local Insight uses 2024 Generalised Clipped (BGC) Ward Boundaries, which can be downloaded from the Open Geography Portal.
For England and Wales, we use the TS007A Census table – Age by five-year age bands to get the population figures for our aggregation and apportioning methodology. This is a more granular measure of population that we also use to create different population age band indicators.
For Scotland, we use the UV101b – Usual resident population by sex by age to get the same information as above.
The ONS publishes detailed methodology papers on best-fit geography and the use of population-weighted centroids. These documents are available in the ONS open geography portal. One useful overview is available from the ONS at this link.
For a guided tour of the Local Insight platform and an opportunity to explore this data for the Wards and Local Authorities you work with, book a demo today.
You don’t need to spend hours making ward profile reports. If you…
More
Local Insight’s dashboard gives you an at-a-glance view of the latest data…
More
The Local Insight Datastore API allows you to connect any of the…
More
