About Nesta

Nesta is an innovation foundation. For us, innovation means turning bold ideas into reality and changing lives for the better. We use our expertise, skills and funding in areas where there are big challenges facing society.

Understanding how flats are heated in England and Wales

Roughly 6.2 million households, or almost a quarter of homes in Great Britain, are living in flats. Decarbonising heating in these homes is crucial for reaching our net-zero targets, but it presents unique challenges. This is due to the need for coordinated action among multiple residents and landlords, the technical complexities of limited space and retrofitting shared systems, and existing regulations such as permitted development rules. 

This is why understanding how flats are currently heated is vital. Our recent analysis dives into this complex landscape, providing a clearer picture of heating systems in flats across England and Wales.

What did we find?

There are about 6.2 million flats in Great Britain according to the EPC data, 5.6 million of which are in England and Wales. Grouping this data by the number of storeys and rise type helps us understand how different types of flats are heated. We used a model to estimate the building storey count for flats from building height and property density data where available (see the methodology section below for more detail). This gave us building storey count estimates for 4.6 million flats.

According to our estimates, low-rise flats (1-3 storeys) are the most common type, comprising at least 46% of flats in England and Wales. 30% of flats are in medium-rise buildings (4-10 storeys), and at least 5% of flats are in high-rise buildings with 11 or more storeys. The remaining 19% of flats are of unknown rise type.

Across all flats, at least 14% use communal heating systems, which refers to a single shared heating system which provides heat to multiple properties, like a communal heat pump or boiler. This is opposed to individual systems which heat only a single property. Communal heating systems appear to be more common in high-rise flats than low- and medium-rise ones, with 45% of high-rise flats using communal heating, compared to 16% and 7% of medium- and low-rise flats, respectively. At least half of all communal heating systems are fossil-fuel based - most of these using gas. The number could potentially be higher because we are missing fuel type data for 44% of communal heating systems.

Conversely, at least two-thirds of all flats in England and Wales use individual heating systems. Of these, 75% of low-rise flats use fossil-fuel heating systems to heat their homes, compared to around 60% of medium-rise flats and 40% of high-rise flats. Therefore, as building rise increases, so does the proportion of flats using electric heating, with around 60% of flats in high-rise buildings using electricity to heat their homes (Figure 1).

If we take a more granular view of flat heating systems (both communal and individual) and look at how they change with the number of storeys in a building rather than rise type, we can see that, up to 10 storeys, as the number of storeys of a building increases, the proportion of flats using electric heating increases and the proportion of flats using fossil-fuel-based heating systems decreases (Figure 2). However, beyond 10 storeys, the proportion of flats with unknown heating system fuel type exceeds 20% and obscures the trend. This missing data is possibly due to difficulties with EPC assessors accessing shared heating systems, which would be an issue in higher rise buildings in particular because, as seen above, a higher proportion of them have communal heating systems.

Since we have fuel type information for all flats with individual heating systems, we have a more complete picture of the trends in the fuel types they use. Looking at these flats specifically, we can see a general upwards trend of increasing proportions of flats on electric heating systems as the number of storeys increases (Figure 3). We see some spikes of fossil-fuel heating at particular storey counts, for example at 29-30 storeys. This could be due to a single high-rise building (or a small number of high-rises) in our dataset in which all flats are on individual fossil-fuel based heating systems dominating the data at these particular storey counts.

What do our findings mean for decarbonising flats?

Our analysis paints a detailed picture of heating systems in flats across England and Wales. We've found that communal heating is far more prevalent in high-rise buildings, used by 45% of flats in these blocks, with at least half of these systems relying on fossil fuels. This is an important insight, as upgrading these large communal systems could decarbonise many homes simultaneously. Conversely, individual fossil-fuel heating dominates in low-rise flats, with 75% using such systems, while electric heating becomes increasingly common in taller buildings, with around 60% of high-rise flats using electricity for individual heating.

This analysis shows that decarbonising flats requires a nuanced, multifaceted approach. The findings underscore the need for distinct strategies for low-rise flats, which predominantly use individual fossil-fuel systems, compared to high-rise flats, where communal systems (often fossil-fuel based) present opportunities for larger scale upgrades. 

Furthermore, this analysis could help us explore complex issues such as the potential cumulative sound effects of heat pumps in different flat archetypes. This is particularly of note for households living in flats, as current planning rules designed to mitigate adverse sound effects often require them to navigate complex planning processes to decarbonise.

For this analysis, we used domestic energy performance certificate (EPC) records up to October 2024 as our baseline dataset, which contains 5.6 million records of individual flats in England and Wales. The 2021 census records 5.4 million households living in flats, so our starting dataset appears to have good coverage of the true number of flats.

For flats, the EPC records the number of storeys in the apartment block. However, this field only contains data for around 8% of all EPC records for flats in England and Wales, so first we had to fill in the missing data for the other 92%. We did this by training a model to predict storey counts for buildings based on two building characteristics: 

  1. the estimated height of the building; and 
  2. the property density of the building, ie, how many flats there are per metres squared of building footprint.

We used building height and footprint data from the Microsoft Global ML Building open dataset, which estimates both building height and footprints from satellite imagery. To calculate property densities, we used OS Open UPRN data to first estimate the number of properties in each building, and then calculated the property density per area of footprint. Height data was available for around 80% of flats in our dataset.

Ultimately, we trained the model on about 343,000 records and then used it to estimate storey count for about 4.2 million flats. We then grouped flats into three different building rise classes depending on how many storeys their apartment block has:

  • 1-3 storeys is classed as low-rise.
  • 4-10 storeys is classed as medium-rise.
  • 11 or more storeys is classed as high-rise.

Our model accurately predicts building rise class about 80% of the time. On average, our model predicts storey counts about 1.4 storeys off the true count, although accuracy declines significantly on buildings over around 23 metres tall and below 5 metres. We know there are underlying data quality issues with the data we train our model on which affects the accuracy of our results:

  • Microsoft’s building height estimation model appears to predict a low number of tall buildings and potentially under-estimates the height of some taller buildings. We speculate that this could be due to variable heights on the same building (Microsoft records average height), among other issues. We have tried to correct for some of this by combining additional open-source data for tall buildings from Wikipedia, but storey count is still likely to be underestimated for some buildings.
  • Multiple buildings can sometimes be combined into one footprint and this will reduce the accuracy of property density and height data.
  • Height and building footprint data is estimated from satellite images taken between 2014 and 2024 so may be out of date for some areas.
  • Height data doesn’t account for basement storeys - and therefore neither does our model, meaning we could underestimate storey count for buildings with basements.
  • The EPC data is known to be erroneous. We have trained our model under the assumption that the storey counts provided in the EPC data are accurate but there may be errors in these data. We excluded Scotland from this analysis due to an apparent systemic error in recording storey count accurately.
  • Although we have significantly reduced the number of missing building rise class values, 18% of flat records are still missing this data.

Author

Roisín Gorman

Roisín Gorman

Roisín Gorman

Data Scientist, Data Science Practice

Roisín works as a data scientist embedded in the sustainable future mission.

View profile
Codrina Cretu

Codrina Cretu

Codrina Cretu

Mission Manager, sustainable future mission

Codrina is mission manager for the sustainable future mission.

View profile
Aidan Kelly

Aidan Kelly

Aidan Kelly

Junior Data Scientist, Data Science Practice

Aidan is a junior data scientist in the Data Science Practice, embedded in the sustainable future mission to focus on the reduction of carbon emissions from UK households.

View profile