Official Statistics

English Housing Survey: local authority housing stock condition modelling, 2019

Published 1 June 2023

Applies to England

1. In this release:

This new Experimental Official Statistics series provides – for the first time – sub-regional estimates of housing stock condition. Specifically, modelled estimates of:

• the number and proportion of occupied homes that are deemed non-decent according to the Decent Homes Standard in each local authority, by tenure and dwelling type (see Decent Homes Standard spreadsheet)
• the number and proportion of occupied homes that are deemed unsafe due to having a Housing Health and Safety Rating System (HHSRS) Category 1 hazard in each local authority, by tenure and dwelling type (see HHSRS Category 1 hazards spreadsheet)

Approximately 16.7% of occupied properties in England were non-decent in 2019, and 9.9% had a HHSRS Category 1 hazard. However, the data suggests that there is wide variation in housing condition across England – ranging from 9% of total occupied dwellings in some regions being non-decent, to more than 30% in others. Similarly, occupied dwellings with Category 1 hazards range from just over 3% in some local authority areas to more than 30% in others.

The models used to produce these estimates were developed using the 2019 English Housing Survey (EHS) housing stock data. It uses household and dwelling characteristics known to be associated with housing quality at a national level to predict housing quality for each local authority in England. More detail on the modelling methodology is provided later in this release.

These statistics have been developed in response to a growing interest in housing quality and, in particular, in conjunction with the levelling up missions of the department. We will continue to develop these metrics to support the needs of users and welcome any feedback on this release.

2. Decent Homes Standard

For a dwelling to be considered ‘decent’ under the Decent Homes Standard, it must:

• meet the statutory minimum standard for housing (the Housing Health and Safety Rating System, since April 2006), homes which contain a Category 1 hazard under the HHSRS are considered non-decent
• provide a reasonable degree of thermal comfort
• be in a reasonable state of repair
• have reasonably modern facilities and services

The modelling estimates there is a great deal of variation across local authorities – from 9% of occupied dwellings in some local authorities, to more than 30% of dwellings in others failing the decent homes standard.

Across all tenures, higher proportions of non-decent occupied dwellings are more likely to be found in local authorities in the North and South West, as well as some districts in the West Midlands.

Proportion of non-decent homes by local authority area

Number of non-decent homes by local authority area

In terms of the number of non-decent occupied dwellings, these estimates follow a similar broad pattern though, predictably, urban areas with higher populations and a greater proportion of rented stock are more likely to have non-decent dwellings.

Non-decent homes by tenure

High proportions of non-decent social and private rented accommodation are not necessarily found in the same local authority areas. There are high proportions of both non-decent social and private rented dwellings in the South West, whereas the Welsh borders are more likely to have a high proportion of non-decent social rented homes, and the North, parts of the East and Yorkshire and Humber are more likely to have a higher proportion of non-decent private rented homes.

Overall, the proportion of non-decent homes varies by tenure. Rented dwellings, particularly those in the private rented sector, are more likely to be non-decent overall.

The below chart shows the distribution of non-decent properties by tenure, across the nine regions in England. The figure indicates that the proportion of non-decent private rented homes is consistently higher than other tenures across nearly all the local authority areas in England. Proportions of non-decent owner occupied dwellings are generally lower than for private rented dwellings, though generally higher than social rented dwellings. However, proportions of social rented dwellings see particularly high peaks in some local authority areas – particularly in the South West and parts of the West Midlands.

Distribution of the proportion of non-decent homes by tenure and region

^{Source: LA stock condition modelling of non-decent homes, 2019-20}

^{Note: Estimates for the Isles of Scilly (South West) are based on small sample sizes and should be treated with caution.}

Drivers of non-decent homes

While this release does not tell us what is driving non-decency in each local area, we do have a sense of the kinds of dwellings that are more likely to be non-decent from the English Housing Survey. The Housing Quality and Condition Report, 2020, finds a strong relationship between both energy efficiency and housing quality – 96% of F and G properties are non-decent – along with age – 32% of properties built before 1919 are non-decent.

Indeed, we can see there is a relationship between the proportion of non-decent occupied dwellings in local authorities, and the proportion of dwellings that were built before 1919.

Relationship between proportion of dwellings built pre-1919 and proportion of non-decent homes

^{Source of dwelling age data: Valuation Office Agency, Council Tax Stock of Properties, England, 2020, LA stock condition modelling of non-decent homes, 2019-20}

Some types of occupied dwellings are also more likely to be non-decent. The modelling found that, across England, 19% of occupied flats and 18% of terraced houses were non-decent, compared to 13% of detached houses, 13% of bungalows, and 16% of semi-detached houses.

3. HHSRS Category 1 hazards

The HHSRS is a risk-based assessment that identifies hazards in dwellings and evaluates their potential effects on the health and safety of occupants and their visitors, particularly vulnerable people. The most serious hazards are called Category 1 hazards and where these exist in a home, it fails to meet the statutory minimum standard for housing in England.

Given that homes that contain a Category 1 hazard also fail to meet the Decent Homes Standard, it is unsurprising that the overall pattern of unsafe dwellings is similar to the pattern of non-decent dwellings.

In terms of dwelling type, Category 1 hazards are more likely to be found in houses than in flats. Across England, 12% of terraced houses have a Category 1 hazard, along with 10% of semi-detached houses, 10% of detached houses, and 9% of bungalows, compared to 8% of flats.

4. Experimental Official Statistics

Experimental Official Statistics are defined in the Code of Practice for Statistics as “new official statistics undergoing evaluation”. They are published to involve users and stakeholders in their development and to build in quality at an early stage.

These statistics are modelled from the English Housing Survey, which provides robust regional and national data (based on physical inspections of properties, undertaken by a qualified surveyor). Application of the model requires specific local area data on a variety of demographic and socio-economic factors. Ideally, EHS data would be used directly to model at the local level. However, this is not possible, given the relatively small sample used for the survey, which does not give sufficient coverage for each of the 309 local authorities. Instead, EHS data has been used to create a model which can then be applied to a national, dwelling level dataset (which includes the 2011 Census small area datasets and 2020 Experian dwelling-level data – more detail on these datasets is available later in this release).

With this approach, it is possible that small areas which are atypical in condition, may not be identified by the model and, where possible, we would encourage data users to compare the modelled results to other local data that might be available.

We welcome feedback on the data, especially where the modelled estimates do not align with other local authority level data and/or intelligence on housing quality. Please send your comments to: EHS@levellingup,gov.uk.

5. Definitions

Decent Homes Standard

For a dwelling to be considered “decent” under the Decent Home Standard, it must:

• Meet the statutory minimum standard for housing (the Housing Health and Safety Rating System (HHSRS). Homes which contain a Category 1 hazard under the HHSRS are considered non-decent
• Provide a reasonable degree of thermal comfort
• Be in a reasonable state of repair
• Have reasonably modern facilities and services

HHSRS Category 1 hazards

The Housing Health and Safety Rating System (HHSRS) is a risk-based assessment that identifies hazards in dwellings and evaluates their potential effects on the health and safety of occupants and their visitors, particularly vulnerable people. The most serious hazards are called Category 1 hazards and where these exist in a home, it fails to meet the statutory minimum standard for housing in England.

6. Future developments

The Department will consider how these statistics should be developed and updated going forward, including in light of feedback received on this statistical release. Our assumption at this stage is that they will be published annually.

The government is committed to levelling up the United Kingdom and has set out a series of priorities and missions to do so. In particular: by 2030, renters will have a secure path to ownership with the number of first-time buyers increasing in all areas; and the government’s ambition is for the number of non-decent rented homes to have fallen by 50%, with the biggest improvements in the lowest-performing areas. Part of our development work will consider how well these experimental statistics can be used to measure such progress.

The current estimates set out in this publication are based on 2011 Census data. One area of development will be to update them using 2021 Census data when this is available.

If you have views on these or other options for future development, please contact: EHS@levellingup,gov.uk.

7. How the data should be used

How the data should be used

• To show local authorities where housing quality might be poorer/worse than average.
• To examine how housing quality might vary by tenure and dwelling type
• To indicate the likely scale of the “problem”, i.e., by estimating the number of non-decent homes in each local authority
• To use alongside other local level data and/or intelligence to develop an understanding of housing condition.

How the data should not be used

• To identify vacant homes in poor condition (only occupied homes are included)
• As a replacement for survey data at the local authority level (if available)
• To estimate all dwelling stock by tenure at a local authority level
• To compare effectiveness of local authority enforcement performance

8. Modelling methodology

The local authority level stock modelling is undertaken by BRE Group’s Local Government Data & Insights team, on behalf of DLUHC. BRE provide local authorities with key housing condition and energy efficiency variables from their housing stock models to enable a targeted intervention approach to improving dwellings and the health of occupiers.

The analysis uses known patterns of dwelling and household characteristics that are related to or associated with the safety and decency of the dwelling to model decency and safety at a local authority level. While we are not able to undertake inspections of decency or safety in a robust way in each local authority, we can use the known characteristics associated with non-decency or unsafe dwellings to estimate the number and of homes in an area that would likely fail the decent homes standard or a HHSRS assessment.

Decent Homes model

The 2019 sub-regional Decent Homes model is based on the English Housing Survey (EHS) combined 2018 and 2019 dataset (referred to throughout as the 2019 EHS data). The EHS is a continuous national survey that collects information about people’s housing circumstances and the condition and energy efficiency of housing in England. It provides robust regional and national data, based on physical inspections of properties, undertaken by a qualified surveyor.

Application of the model requires specific local area data on a variety of demographic and socio-economic factors. The key datasets contributing to the modelling process were:

• 2019 English Housing Survey data
• 2020 Experian dwelling-level data
• EPC data to September 2021
• 2011 Census demographic data

Ideally, EHS data would be used directly to model at the local level. However, this is not possible, because of the size of the sample used for the survey, which does not give sufficient coverage for each of the 309 local authorities. To produce results at this level, EHS data has been used to create a model which can then be applied to a national (dwelling level) dataset. With this approach, it is possible that small areas which are atypical in condition, may not be identified by the model and, where possible, we would encourage data users to compare the modelled results to local data.

The modelling process can be summarised in the following 5 stages:

1. Regression model testing

A large number of independent variables are individually tested against the binary Decent Homes indicator to determine which have the highest coefficient of determination, i.e., which are most sensitive in predicting the Decent Homes outcome. The variables are then grouped according to this coefficient and an iterative selection is made by running logistic regression on a developing group of these variables, with those least likely to contribute to final model predictions being dropped at each stage. The final set of variables contributing to the model is dependent on the results of this regression testing.

The 2019 decent homes model is based on the following eight variables, derived from the public and commercial sources listed above:

• Dwelling age – pre-1919, 1919 to 1944, 1945 to 1980, post-1980 (Experian ConsumerView data)
• Mosaic household classifications by postcode – 15 socio-economic groups as detailed in the Mosaic guide (Experian Segmentation data)
• Proportion of central-heated dwellings per census output area (COA) (2011 Census)
• Tenure – owner occupied, private rented, social rented (Experian ConsumerView)
• Proportion of single person, non-pensioner households per COA (2011 Census)
• Typical number of bedrooms by postcode area (Experian ConsumerView)
• Dwelling type – terraced, detached, semi-detached, bungalow, flat (Experian ConsumerView)
• Predominant age of dwelling per postcode area (Experian ConsumerView)

In addition to the regression modelling, the national set of lodged EPC data covering around 60% of the dwelling-level dataset is used to model the thermal comfort criterion as closely as possible.

2. Model application

The coefficients for each category of the final model’s independent variables are run to give the predicted likelihood for each dwelling in each area to be non-decent. A further coefficient is included where the EPC data predicted that a dwelling fell below the thermal comfort measure.

These likelihoods are then combined to produce the total number and proportion of non-Decent dwellings at COA level. This is run separately for each region using a multiplying factor calculated from the modelled and published regional Decent Homes proportions applied to each COA non-decent total, thus ensuring that the non-decent regional proportions from the model are consistent with the EHS analysis.

3. Output area cloning

Due to geographical changes between the 2011 Census and other national data used in the modelling, a small number of COAs do not have Decent Homes data. These gaps are filled by identifying COAs with complete data, which shared identical or very similar characteristics to the missing cases. Census data is used for this comparison as it records proportions of particular characteristics rather than discrete categories, and so the process can more accurately match pairs of COAs.

4. Consolidation to national figures

To provide consistency between the modelled dwelling and Decent Homes numbers and the EHS regional figures, an adjustment is made in which small incremental changes are assigned randomly at COA level, until the totals of all dwellings and non-decent dwellings are consistent with the published regional totals. The model assumes that any inconsistencies are randomly distributed.

5. Validation and aggregation

The adjusted figures are re-checked against the published totals before the data is aggregated to local authority and regional level.

Further validation is applied to ensure the proportion of non-decent homes in each COA is within an expected range, and that regional totals are consistent at each level of aggregation.

Finally, the dwelling totals and number of non-decent homes at local authority level are consolidated to ensure consistency with both EHS regional totals and 2019 local authority totals drawn from DLUHC data on the number of dwellings by tenure and district (Table 100). The published set of non-decent proportions are calculated from these consolidated dwelling totals. In practice this means that the modelled estimate of the number of non-decent homes in each region is the same as the total number of occupied non-decent homes observed in the English Housing Survey.

HHSRS Category 1 hazard model

The same modelling approach is taken for the HHSRS Category 1 hazard modelling, with EPC data being used in this model to estimate the Excess Cold hazard and a slightly different set of key variables contributing after regression testing. The HHSRS model was based on coefficients calculated for the following variables:

• Dwelling age – pre-1919, 1919 to 1944, 1945 to 1980, post-1980 (Experian ConsumerView data)
• Mosaic household classifications by postcode – 15 socio-economic groups as detailed in the Mosaic guide (Experian Segmentation data)
• Tenure – owner occupied, private rented, social rented (Experian ConsumerView)
• Proportion of central-heated dwellings per census output area (COA) (2011 Census)
• Government Office Region
• Typical number of bedrooms by postcode area (Experian ConsumerView)
• Households with or without children (Experian ConsumerView)
• Proportion of converted flats per COA (2011 Census)
• Predominant age of dwelling per postcode area (Experian ConsumerView)

9. Related data

English Housing Survey

The English Housing Survey is a continuous national survey commissioned by the Department for Levelling Up, Housing and Communities (DLUHC). It collects information about people’s housing circumstances and the condition and energy efficiency of housing in England. It provides robust regional and national data on housing quality, based on physical inspections of properties, undertaken by a qualified surveyor. National Statistics publications from the English Housing Survey, on housing quality can be found here:

• Data on dwelling condition and safety
• Housing quality and condition report

Energy Performance Certificate (EPC) data

Based on certificates held on the Energy Performance of Buildings Register, these Energy Performance of Buildings Certificates statistics present information about the energy efficiency of domestic and non-domestic buildings in England and Wales that have been constructed, sold, or let since 2008, and of larger public authority buildings recorded since 2008. For the modelled estimates of housing quality, this data was used to model the thermal comfort criteria of the Decent Homes Standard, and the Excess Cold hazard in the HHSRS Category 1 hazard model.

There is a strong relationship between energy efficiency and housing quality, and most homes with poor energy efficiency do not meet the Decent Homes Standard.

Experian data

The modelling is supported by Experian household level data (PDF, 6.69MB) which provides an up-to-date portfolio of address, geographic and demographic data, which is used to inform BRE’s stock modelling estimates. Information on dwelling age, household composition, tenure and dwelling type have informed the likelihoods used in the Decent Homes and HHSRS Category 1 hazard models and helped to estimate the number of dwellings failing to meet each indicator.

2011 Census demographic data

During the regression analysis stage of this modelling, the 2011 Census data at COA level, was used to help identify demographic categories that were most likely to contribute to estimates of these housing quality indicators. The data were further used to provide a baseline position for the estimated number of dwellings in each local area, which were then refined through consolidation with recent data from EHS and other central government sources. When available, 2021 Census data will be used to update the models.

Welsh Housing Conditions Survey

The Welsh Housing Conditions Survey collects information about the condition and energy efficiency of all types of housing in Wales.

Scottish House Condition Survey

The Scottish House Condition Survey contains statistics on house condition, based on an annual survey sample of around 3,000 dwellings.

Northern Ireland Housing Statistics

The Northern Ireland Housing Statistics report contains information and statistics relating to the condition of homes in Northern Ireland.

10. Technical notes

Actual numbers of dwellings by local authority, tenure and dwelling type are indicative and based on the outputs from the modelling methodology described above. As such, stock totals and the numbers of non-decent homes, whether explicitly stated or inferred, should be viewed as an approximate measure of relative size against other local authorities, but should not be treated as surveyed totals.

The proportions of homes which are non-decent have been consistently modelled across the local areas as described above, and are indicative of their relative levels of non-decency, but are not based on a surveyed total of non-decent homes.