Welsh Index of Multiple Deprivation (WIMD) 2025 technical report - Housing domain

Conceptually, the purpose of the housing domain is to identify inadequate housing, in terms of physical and living conditions and availability. Here, living condition means the suitability of the housing for its inhabitant(s), for example in terms of health and safety, and necessary adaptations. The domain has a relative weight of 9% in the overall index.

Conceptually, the purpose of the housing domain is to identify inadequate housing, in terms of physical and living conditions and availability. Here, living condition means the suitability of the housing for its inhabitant(s), for example in terms of health and safety, and necessary adaptations. The domain has a relative weight of 9% in the overall index.

Indicators

The housing domain has two equally weighted sub-domains, housing availability and housing conditions, with two equally weighted indicators each. 

The housing availability sub-domain contains indicators of:

• overcrowding: the percentage of households that are overcrowded (bedroom measure)
• inability to afford to enter owner occupation or the private rental market

The housing conditions sub-domain contains indicators of:

• poor quality: the likelihood of housing being in disrepair or containing serious hazards (for example, risk of falls or cold housing)
• energy efficiency

Overcrowding 

Type of indicator

Percentage.

Numerator

Number of households that are overcrowded (bedroom measure).

Denominator 

Number of households.

Source and time period 

2021 Census, Office for National Statistics (ONS).

Additional notes

This indicator provides a measure of whether a household is overcrowded. The definitions of overcrowding are fully explained on the ONS website (Census 2021 metadata).

Comparability with WIMD 2019

Not directly comparable.

WIMD 2025 measures overcrowding as the percentage of households that are overcrowded, where as WIMD 2019 it was measure as the percentage of people in overcrowded households. 

Updated data are available based on Census 2021, however the breakdowns needed for WIMD are only available on a household basis and no longer on a resident basis, i.e. there is data on the percentage of households that are overcrowded, rather than the percentage of people in overcrowded households. 

We have compared overcrowding data for LSOAs on a household vs resident basis using Census 2011 outputs (which include both definitions) and found there to be a high level of correlation, so changing between these definitions has little impact on ranks for the housing domain.

This change to the indicator has the advantages of increased transparency and accessibility (being based on data already published by ONS), comparability with the approach for England, and potential for added value in terms of allowing users to access additional breakdowns of the indicator data from Census 2021 custom tables on the ONS website.

Inability to afford to enter owner occupation or the private rental market 

Type of indicator

Overall indicator is a score (indicator sub-components are rates, see additional notes).

Numerator

For the indicator sub-components: the number of relevant households estimated to be unable to afford to enter owner occupation or the private rental market for the relevant cohort.

Denominator

For the indicator sub-components: the total number of relevant households.

Source and time period 

The main data sources are:

• the Family Resources Survey (FRS) for household incomes and composition, financial years ending 2024, 2023, 2022 and 2020
• the ONS House Price Index (formerly Land Registry) for house prices, 2023
• data from Rent Officers Wales (equivalent to administrative data collected in England by the Valuation Office Agency) for rents, 2023 

Other sources used included a range of Census 2021 and other published or official data at LSOA level, including ONS 2021-based classifications of LSOAs and local authorities, ONS populations for LSOAs, NOMIS claimant counts, and some other indicators at local authority level including from the Annual Population Survey and the Annual Survey of Hours and Earnings.

Private rent data from Rent Officers Wales reflects achieved rents across tenancies of varying lengths. This may not accurately represent the cost of securing a new tenancy today, as advertised rental prices are dictated by the current market conditions, which are generally higher than existing tenancies.

Additional notes

 We have introduced a new indicator on inability to afford to enter owner occupation or the private rental market produced by Heriot-Watt University. This indicator has been used in several iterations of the English indices of deprivation (MHCLG), for younger households with head of household aged under 40. Our measure also considers older private renters with a head of household aged 40 to 65. 

The indicator is modelled in several stages, starting with microdata from the FRS, then applying predictive functions to Census and other small-area data, and finally controlling for consistency at the Wales level. 

Affordability is assessed using two main criteria: the ratio of lower quartile house price or rent (by bedroom size) to income, and the ratio of residual income after housing costs to a standard based on basic household requirements. To derive house price and rent thresholds relevant to a given household, the approach uses housing market areas defined via research into commuting and migration patterns, recognising that people search for housing across larger geographies. 

The overall indicator score is based on 4 sub-components, which are estimated proportions of:

• difficulty of access to owner-occupation for households where the head is aged under 40
• difficulty of access to the private rental market where the household head is under 40
• difficulty of access to owner-occupation for older private renters (with household head aged 40 to 65)
• difficulty of access to the private rental market for older private renters (with household head aged 40 to 65)

We have published the data for each component (by age and tenure) separately as part of the WIMD indicator datasets. These may be more meaningful to users than the overall indicator score which is calculated as follows from the 4 sub-components:

• for the 2 age groups separately, the proportions unable to buy or rent are standardised and combined with equal weight
• these 2 measures are combined into an overall indicator score by standardising and combining them with 75% weighting for the younger group and 25% for older private renters

Please see annex 7.1 for further details.

Comparability with WIMD 2019

New indicator.

Likelihood of poor quality housing (being in disrepair or containing serious hazards)

Type of indicator

Percentage.

Numerator

Estimated number of dwellings that: 

• contain a category 1 hazard for excess cold, falls or other hazards under the Housing Health and Safety Rating System (HHSRS)
• or are in a state of disrepair

Denominator 

Numbers of residential dwellings. 

Source and time period 

Building Research Establishment, using various survey and administrative data sources, 2023.

Additional notes

A dwelling is determined to have a Category 1 hazard as a result of excess cold if there is a severe threat from sub-optimal indoor temperatures. A dwelling is assessed as having a Category 1 hazard in terms of falls if there is determined to be a serious risk of falling on stairs, between levels, level surfaces or falling associated with a bath, shower or similar facility. A dwelling is said to be in disrepair if at least one of the key building components is old and needs replacing or major repair due to its condition; or more than one of the other building components are old and need replacing or major repair. Further details on the modelling process are provided in annex 7.2.

Comparability with WIMD 2019

The update was based on a similar approach to that used for 2019 WIMD and includes the use of more recent data sets where available. The 2025 modelling work also includes the integration of two data sets provided by the Welsh Government; Rent Smart Wales and Welsh Housing Quality Standard (WHQS) to help identify private rented and social stock, respectively.

For the 2019 indicators, data from the Welsh Housing Condition Survey (WHCS) was used to benchmark the modelled data; however, there has not yet been an update to the WHCS. Instead, notional 2023 benchmarks were estimated by comparing changes in each indicator for the West Midlands region between 2017 and 2023 and applying this change to the WHCS 2017 totals to create adjusted likelihoods.

Energy efficiency 

Type of indicator

Average score.

Numerator

Total value of final (observed and imputed) energy performance certificate (EPC) scores.

Denominator 

Total number of residential properties.

Source and time period 

Numerator: EPC data for assessments undertaken between January 2012 and December 2024 (MHCLG open data).

Denominator: Ordnance Survey National Geographic Database.

Additional notes

We have introduced a new measure of energy efficiency, based on average Standard Assessment Procedure (SAP) scores for residential properties in the area, using data from Energy Performance Certificate (EPC) records. MHCLG have developed this indicator for residential dwellings in England and Wales, imputing estimated EPC scores for all dwellings that do not currently have a valid EPC. 

Since around 55% of properties in Wales do not have a valid EPC, the measure imputes estimated EPC scores for all dwellings that do not currently have a valid EPC. EPC open data was linked to Ordnance Survey data including building-level characteristics (such as building type and age) to inform the imputation, based on an average (median) EPC score for nearby properties of similar type, as described below.  

Nearest-neighbour imputation methodology process:

• only EPC records from 2012 onwards were used, as data quality prior to this date was less reliable
• each property without an EPC was matched to similar dwellings (based on dwelling type, build period, and construction material) within its own or directly contiguous LSOAs
• an EPC score was estimated for each property using the mean score of its five closest neighbouring properties of similar type (generating estimates for 98% of properties lacking valid EPC ratings)
• for unresolved properties, the imputation was re-run using relaxed matching criteria: first removing construction material as a criterion, then allowing matches across the local authority

Before calculating mean (of observed and imputed) EPC scores for each LSOA, a final adjustment was made to address potential distortion caused by holiday caravan parks. In affected LSOAs, the number of caravans (and their associated scores) were scaled to align with Census 2021 counts of households resident in caravan accommodation.

Please see the English indices of deprivation 2025 technical report (MHCLG) for further details on the approach. 

Separately, the WG data science unit (DSU) have used machine learning to build a virtual EPC dataset for all homes in Wales which has been used to quality assure the data used in WIMD. We have used the simpler MHCLG model for WIMD 2025 as this will provide consistency with the approach for England, and since the impact of the choice of data on WIMD was small. This choice of data source will be reviewed ahead of the next index as the DSU project develops.

Comparability with WIMD 2019

New indicator.

There are 4 indicators in the housing domain, split into two sub-domains and weighted as follows.

• Housing availability.
  • Overcrowding 25%.
  • Inability to afford to enter owner occupation or the private rental market 25%.
• Housing conditions.
  • Poor quality 25%.
  • Energy efficiency 25%.

The relevant indicators within each of the sub-domains are standardised by ranking and transforming to a normal distribution and combined using equal weights. 

Sub-domain scores are then standardised through being ranked, which gives us the sub-domain ranks for availability and conditions. These two sets of ranks are then transformed to exponential distributions and combined with equal weights to create the overall domain score.

The domain has a relative weight of 9% in the overall index. This has increased from 7% in the 2019 index, due to the addition of new indicators on energy efficiency and the inability to afford suitable housing.

Changes since WIMD 2019 

In WIMD 2025, two new indicators have been introduced into the housing domain: 

• inability to afford to enter owner occupation or the private rental market (part of the housing availability sub-domain)
• energy efficiency (part of the housing conditions sub-domain)

These additions reflect feedback received during the 2025 WIMD proposals survey, where respondents supported their inclusion to better capture housing deprivation. 

The domain continues to include an indicator on overcrowding, now based on Census 2021, which captures aspects of both housing availability and living conditions. As explained in a previous section, the indicator now reflects the percentage of households that are overcrowded, rather than the percentage of people living in overcrowded households (used for WIMD 2019).

As a result of the indicator changes that better capture housing deprivation, the housing domain has been given an increased weight in the overall index in WIMD 2025. The domain now has a relative weight of 9% in the overall index, increasing from 7% in the 2019 index.

Additional information

Due to a lack of robust alternatives at small area level, modelled data are now used in 3 of the 4 indicators in the housing domain. Our quality assurance process suggests that relative trends shown in the data remain reasonably accurate. However, modelled data may have limitations in reflecting the impact of recent housing interventions or other changes. Where decisions are being informed, modelled data should be used alongside robust, up-to-date local intelligence wherever possible.

In addition to domain ranks and indicator values, we have published housing sub-domain ranks on StatsWales.

Overview

A new indicator developed for WIMD 2025 by Heriot-Watt University measures the difficulty of accessing owner-occupation or the private rental market. It primarily focuses on younger households (with a head aged under 40), assessing their ability to afford to buy or rent a suitably sized home at local threshold prices and rents in 2023. The indicator incorporates a secondary element for older private renters (aged 40 to 65), reflecting the growing vulnerability of this group. 

These indicators are comparable to those provided to the Ministry of Housing Communities and Local Government (MHCLG) to update the English indices of deprivation. 

Indicator description

The indicator is based primarily on the estimated proportions of younger (aged under 40) households able to:

• afford to buy a home of appropriate size (based on household composition) at the local threshold price level in 2023
• afford to rent a home of appropriate size in the private market at the local threshold rent level in 2023

This aims to capture the cohort of households entering the housing market, recognising that most first-time buyers are in the younger adult age group. 

The indicator also considers an additional group in potential need, by including estimates for the proportion of ‘older private renters’ (with household head aged 40 to 65) able to afford buying or market renting in the local area. This group has grown in size and remains relatively more vulnerable in terms of affordability, security and dwelling quality and hence potential candidates for more affordable alternatives.

The rationale for focusing on these groups is that they are the main target for local housing and planning policies for the provision of additional social and affordable housing.

All indicators are model-based estimates available down to the lower super output area (LSOA) level. The methodology broadly follows that used in the 2015 and 2019 indices of deprivation for England, but with some detailed methodological changes and the extension to older private renters.

Using these four sets of estimated proportions, we calculate a ‘housing affordability’ measure of the difficulty of access to housing for the two age groups separately, by inverting, standardising and combining the ‘buying’ and ‘renting’ elements with equal weight. These two measures are then combined into an overall housing affordability indicator measuring the inability to afford to enter owner-occupation or the private rental market. This is done by standardising and combining the measures for the two age groups with 75% weighting for the younger group and 25% for older private renters. This reflects the relative size of these populations and arguably provides a reasonable representation of their relative importance for housing and planning policies.

Methods and data sources

Data sources

The main data sources are the Family Resources Survey (FRS) for household incomes and composition, the ONS House Price Index (formerly Land Registry) for house prices, and data from Rent Officers Wales (equivalent to administrative data collected in England by the Valuation Office Agency) for rents. Other sources used included a range of Census 2021 and other published or official data at LSOA level, including ONS 2021-based classifications of LSOAs and local authorities, ONS populations for LSOAs, NOMIS claimant counts, and some other indicators at local authority level including from the Annual Population Survey and the Annual Survey of Hours and Earnings.

Private rent data from Rent Officers Wales reflects achieved rents across tenancies of varying lengths. This may not accurately represent the cost of securing a new tenancy today, as advertised rental prices are dictated by the current market conditions, which are generally higher than existing tenancies.

Income

Income is defined as the income of the ‘first benefit unit’ in the household, excluding income from means-tested benefits and from benefits intended to cover the additional costs associated with disability (DLA, PIP, AA). The first benefit unit is defined as the main householder and any partner where the household reference person is aged under 40. Other adults present in any ‘complex’ household are separate benefit units, and their income is not included because these would not be considered reckonable income for the purposes of obtaining a mortgage and because it is assumed that it is the core benefit unit that would be seeking to buy or rent an appropriate housing unit. This means that ‘concealed households’ (such as adult children or other adults living within a larger household) are not included in the measure, even though they may also be trying to access housing independently.

Housing market areas

As far as assumptions about where a particular family may be looking to live, the approach uses “housing market area” (HMA) geographies rather than LSOAs, as people are likely to reach wider than an LSOA boundary when looking to buy or rent. There would also likely be insufficient price and rent data at small area level to derive reliable threshold costs for different sized accommodation to input to this model.

The threshold house prices and rents for England in preceding work were calculated for HMAs derived from the 2010 study of The Geography of Housing Market Areas in England (Jones, Coombes and Wong, 2010), undertaken for the former National Housing and Planning Advice Unit and published by the Department for Local Government and Communities.

This work aimed to identify the optimal areas within which planning for housing should be carried out, using Census data about commuting and migration patterns. The lower tier of local market areas (LHMAs) was used for this purpose, with small area price data and local authority level rent data apportioned to LHMAs. The same approach is followed for Wales.

Figure 7.1: map of local housing market areas in Wales used in WIMD 2025

Image

Description of figure 7.1: the map shows that the 35 resultant LHMAs are close to the current local authority areas, with subdivisions in several cases. Examination of 2001 ward level information on the LHMAs for Wales suggested that there was only one pair of authorities (Neath-Port Talbot and Bridgend) where there were substantial areas crossing LA boundaries, plus three larger ‘valleys’ type authorities where smaller HMAs representing different valleys were part of the system. In addition three large rural authorities were also split into more than one HMA.

Rental data for local authorities was used to approximate rental data for LHMAs using evidence on the relationship between rents and prices: variation in rents between areas was found to be in proportion to the variation in house prices raised to the power of 0.7. Mapping LHMAs onto LSOAs involved some approximation, since LHMAs were built from wards and some originally crossed the border with England.

The measure does not aim to capture variation in house prices between small areas within a given authority or city, nor does it attempt to capture the availability or supply of suitable housing. Instead, it reflects likely variation in household income and bedroom requirements at the small area level, alongside a threshold housing cost based on the wider housing market area. The aim is to estimate how likely it is that households in a small area can reasonably afford to buy or rent a suitable home within the broader market area. 

While other datasets, such as House price statistics for small areas in England and Wales (ONS), focus on local price variation, this measure takes account of household circumstances and assumes that people may move a reasonable distance to find affordable housing.

For example, a household living in a relatively low-income neighbourhood may not be able to afford housing in their immediate area, but could potentially afford a suitable property in a nearby part of the wider housing market area. The measure reflects this by combining local income and household composition data with housing cost thresholds set at the broader market level, rather than attempting to capture fine-grained price differences between adjacent neighbourhoods.

Affordability thresholds and criteria

Whether for buying or renting, a household must pass two ratio-based affordability thresholds, one relating the ratio of house price or rent to income, and the other based on the ratio of residual net income after housing costs to a standard based on household basic requirements allowing for household composition. Criteria used in previous work were updated to account for academic analyses, market practices, and recent evidence on housing choices of key target groups. 

Price/rent thresholds: The threshold price is based on the lower quartile of all sales within size groups (1, 2, 3 and 4+ bedroom) at Housing Market Area level, and similarly the lower quartile private market rent within size groups. The size criteria for affordability to buy includes a spare bedroom. For affordability to rent in the market the lowest size category included was one-bedroom.

Ratio of price/rent to income: For buying the ratio is a lending multiplier from gross income to maximum mortgage (4.0 for one earner 3.6 for 2 earners), whereas for renters it would be a ratio rent to net income of 30%. 

For house purchase, it is assumed that a 95% mortgage would be taken out but allowance is also made for ‘excess’ savings (over £12,000) already held by the household to be contributed. Some households are enabled to buy thanks to large wealth transfers from parents or other relatives; the indicator makes no allowance for this. The chosen lending multipliers reflect recent practice and are consistent with the ratio of payment to net income of 30% with interest rates at around 4% and mortgage term of 30 years (different terms and multipliers would apply to older renters depending on age).

Ratio of residual income to basic requirement: For both tenures the residual income ratio is set at 1.4 times an amount derived from 90% of selected core budget items from the Minimum Income Standard (MIS).

Modelling sequence

The indicator is modelled using FRS data, as this has a large sample (with pooling of four years of data) and contains the most authoritative income and benefits data. The years of data used were financial years ending 2024, 2023, 2022 and 2020. Data for financial year ending 2021 is not considered sufficiently complete or reliable owing to Covid restrictions on survey fieldwork.

Estimation of incomes and affordability measures occurs in stages. First a model is built using certain variables in the FRS sample survey data to predict the measures of income and affordability. Then a dataset of equivalent predictor variables is created using Census and other sources. The predictive model parameters from the FRS analysis are then applied to generate predicted values for the incomes and affordability measures at LSOA level. These LSOA estimates are then adjusted for consistency using control values based on area types (whether in higher or lower priced regions) and local authority categories, applying an iterative fitting method. 

At the final stage the individual LSOA controlled predicted values were checked for outliers or missing cases. In these cases, an alternative estimate would be generated using ‘affordability elasticities’ (generated within the FRS analysis) applied to price or rent differences from group averages. 

The approach involves modelling incomes and affordability across England and Wales as a whole, with the same predictive functions applying to both as FRS sample numbers for Wales are not adequate to enable effective modelling for Wales alone. However, at the stage of controlling the estimates, predicted affordability rates are checked for consistency with the FRS micro estimates at the Wales level.

Limitations and considerations 

Concealed households

The measure focuses on the income of the ‘first benefit unit’ in a household and excludes other adults in complex households. This means that concealed households (such as adult children or unrelated individuals living within a larger household) are not captured, even though they may face affordability barriers. Most new household formation takes place within the age range up to 40, so it is assumed that this group provides a broad representation of new households entering and becoming established in the housing system.

Private rent data 

The rental data used is sourced at the local authority level, which limits the ability to reflect finer-grained variation in rents within authorities (and some of the LHMAs in use are smaller sub-divisions of local authorities). Reflecting wider evidence, rents are assumed to vary between HMAs within local authorities, proportional to the variation in house prices to a power of 0.7 Additionally, the data may underrepresent smaller landlords and exclude tenancies involving housing benefit, potentially skewing results away from the lower end of the market.

Housing market areas 

Affordability thresholds for each LSOA are based on wider housing market areas (rather than individual neighbourhoods). This reflects the assumption that households may move a reasonable distance to access desired housing, but it means the measure does not capture very localised price differences.

Tenure and age focus 

The indicators target younger households (under 40) and older private renters (aged 40 to 65), based on their relevance to housing policy. However, affordability challenges may also affect other groups not directly covered. The overall indicator does not account for variations in age distribution across areas of Wales, that is, the two indicators are combined into an overall measure using the same weighting for all small areas, 75% weighting for the younger group and 25% for older private renters. However, we will publish the indicator data for each component (by age and tenure) separately, allowing for further analysis if required.

Affordability thresholds

The measure uses two criteria: housing cost-to-income ratio and residual income after housing costs, based on updated evidence and modelling. While robust, these thresholds are still simplifications of complex household circumstances.

Sample size constraints

For older private renters, sample sizes in the FRS are relatively small, which limits the precision of estimates and required further grouping of areas to meet minimum reporting thresholds.

Overview

We commissioned the Building Research Establishment (BRE) to construct a poor quality housing indicator for WIMD 2025, at small area (LSOA) level. BRE used a similar methodology to that they used for WIMD 2019 when the indicator was first created. The indicator is calculated using a model built from survey data, which makes probabilistic predictions about individual level dwellings in Wales, using a range of administrative datasets as inputs.

Indicator description

Conceptually, the purpose of the housing domain is to identify inadequate housing, in terms of physical and living conditions and availability. The poor quality housing indicator estimates the likelihood that dwellings in a given area: 

• contain a Category 1 hazard for excess cold, falls or other hazards under the Housing Health and Safety Rating System (HHSRS)
• or are in a state of disrepair 

These two measures are useful and well-established indicators of housing deprivation, also used for the 2025 update of the English indices of deprivation (MHCLG).

Hazards

The Housing Health and Safety Rating System (HHSRS) is used by Welsh Government as an evidence based risk assessment procedure for residential properties. The HHSRS is a means of identifying defects in dwellings and of evaluating the potential effect of any defects on the health and safety of occupants, visitors, neighbours and passers-by. The system provides a means of rating the seriousness of any hazard so it is possible to differentiate between minor hazards and those where there is an imminent threat of major harm or even death. The emphasis is placed on the potential effect of any defects on the health and safety of occupants, visitors, and particularly vulnerable people. The HHSRS assesses 29 types of hazards. For the purposes of this modelling, hazards were grouped into the following three categories:

1. excess cold

2. falls (comprising of falls on stairs, falls on the level, and falls associated with baths)

3. other (all other hazards)

Dwellings where at least one Category 1 hazard (hazards with potential extreme harm outcome) was present were reported as failing the assessment. A dwelling is determined to have a Category 1 hazard as a result of excess cold if there is a severe threat from sub-optimal indoor temperatures. A dwelling is assessed as having a Category 1 hazard in terms of falls if there is determined to be a serious risk of falling on stairs, between levels, level surfaces or falling associated with a bath, shower or similar facility.

Disrepair

The same disrepair criterion as used in the Decent Homes Standard (England) was used, (the same as that used for the 2025 update of the English indices of deprivation). A dwelling failing the disrepair criterion means that certain building components are in poor condition, defined as either:

1. one or more key building components are old and, because of their condition, need replacing or major repair

2. two or more other building components are old and, because of their condition, need replacement or major repair

Key building components are those which, if in poor condition, could have an immediate impact on the integrity of the building and cause further deterioration in other components. They are the external components plus internal components that have potential safety implications and include:

• external walls
• roof structure and covering
• windows/doors
• chimneys
• central heating boilers
• gas fires
• storage heaters
• plumbing
• electrics

Other building components are those that have a less immediate impact on the integrity of the dwelling. Their combined effect is therefore considered, with a dwelling failing the disrepair standard if two or more elements are old and need replacing, or require immediate major repair.

Methods and data sources

The approach that was used to create the modelled data involved:

• using data from housing condition surveys where experienced surveyors carried out physical inspections of a sample of properties from all tenures (including measurement of disrepair and risks of hazards across all types of dwellings)
• building a model from this data to predict the likelihood of poor quality housing based on a range of predictors (such as the age, type, size, tenure, construction and energy variables such as heating and fuel type)
• applying this model to all dwellings in Wales, using data from a range of sources (including Ordnance Survey, Land Registry, EPC data) to provide the required predictors
• benchmarking the results to national estimates of poor housing quality 

The modelling process above is carried out separately for the aspects of poor quality housing listed below (and defined further in following sub-sections), at a dwelling level:

• the presence of a Category 1 hazard for excess cold (using SAP ratings as a proxy measure)
• the presence of a Category 1 hazard for falls
• the presence of a Category 1 hazard for a hazard other than excess cold or falls
• being in disrepair 

A dwelling is classed as being of poor quality housing if it is predicted to have any of the above features. Estimated data for individual dwellings are then aggregated to LSOA level to provide a rate of dwellings which are classed as poor quality housing for each LSOA.

Developing a housing stock model for Wales 

BRE have developed and used BRE housing stock models for many years. These dwelling level models are used to estimate the likelihood of a particular dwelling meeting the criteria of interest. These outputs can then be mapped to provide a geographical distribution of each of the indicators. The process itself is made up of a variety of data sources, calculations and models. 

The housing stock model developed for Wales consists of the following datasets: 

• Ordnance Survey geographic data
• Experian dwelling-level data
• Xoserve data 

Other datasets were then integrated into this base model to provide enhanced information relating to dwelling tenure and dwelling energy data: 

• Rent Smart Wales data
• Welsh Housing Quality Standard (WHQS)
• Land Registry Commercial and Corporate Ownership Data (CCOD)
• Energy Performance Certificate (EPC) data

The Welsh address list was used as the spine to link the other datasets and to form the base data for the model. The dwelling level data for Wales was then used to produce the model for dwellings in disrepair and dwellings where at least one HHSRS Category 1 hazard was present. 

Determining HHSRS Category 1 hazards 

BRE have developed a method for determining whether a dwelling fails the HHSRS Excess cold hazard: the BRE SimpleCO2 simplified energy model is designed to approximate SAP scores using a minimal level of data. This is an ideal tool for situations where dwellings have not been surveyed in detail but a measure of their energy efficiency is needed. For the remaining HHSRS hazards, BRE’s standard logistic regression analysis was used. 

Determining disrepair 

To determine dwellings in disrepair, logistic regression analysis was used to establish relationships between disrepair and various dwelling and social characteristics. Once these relationships were established, they were used to create a regression model which calculates the probability of a dwelling failing the disrepair criterion. 

Determining the poor housing indicator 

For the 2019 indicators, data from the Welsh Housing Condition Survey (WHCS) was used to benchmark the modelled data; however, there has not yet been an update to the WHCS. Instead, notional 2023 benchmarks were estimated by comparing changes in each indicator for the West Midlands region between 2017 and 2023 and applying this change to the WHCS 2017 totals to create adjusted likelihoods.

A review of the latest English Housing Survey data indicated that regional housing characteristics had not changed significantly since the previous modelling exercise, and as such, the West Midlands was selected as the comparator region for Wales, consistent with the approach taken in the previous analysis.

Once the benchmarked model based on the disrepair and HHSRS components had been created, they were combined to make an overall poor quality housing Indicator. If a dwelling failed one or more of the components, then it failed the poor quality housing Indicator. The dwelling level outputs for Wales were then aggregated to LSOA level.