Nine ‘myths’ about the building stock of Great Britain

2.1 MYTH 1: ‘THE UNIT BY WHICH THE STOCK IS BEST MEASURED IS THE BUILDING’

Despite the name of the new NBD, and the fact that it covers almost all buildings in the country, it is not primarily organised by the building as its basic unit. It is organised first by premises. A premises is an extent of contiguous or adjacent floorspace with a single occupier. Much of the data in the NBD come from the Valuation Office Agency (VOA) of His Majesty’s Revenue and Customs (HMRC), which estimates the values of non-domestic properties for taxation purposes.² Occupiers are then responsible for paying the taxes. The VOA uses the premises (or what it calls the hereditament) as its overall unit of assessment. The VOA also determines council tax bands for domestic (residential) premises.

Imagine a row of 10 terrace houses, all built at the same time in the same style (Figure 2). Is this one building or 10? For most purposes it would be sensible to treat it as 10 buildings, i.e. as 10 houses accommodating 10 households. In this case every building/house equates to a premises. Every house/premises has its own gas and electricity meters. Each household, its activities and the building it occupies can be studied or modelled separately. For owner-occupiers, it is the household that makes decisions about energy use and efficiency measures. If flats are excluded, then the domestic building stock is relatively simple in its relationships of premises to buildings.

Figure 2

Now imagine that the ground floor of one house is converted into an office, and on the upper floors are two flats (Figure 3). This needs to be treated as three premises. The patterns of activity and uses of energy in the flats and the office will be different. Each is likely to have its own electricity meter (although all three might possibly share a gas meter). If the former house is still regarded as one building, without respect to premises, then floor areas devoted to different activities would have to be added together, and energy use averaged. In the NBD, this would by contrast be treated as three premises sharing one building. Importantly, when premises and buildings are distinguished, data on the character of the occupiers and their activities may be attached to the premises, and data on construction, materials and the geometry of the envelope to the building.

Figure 3

As a further level of complexity, imagine a small supermarket takes over the ground floors of two buildings by breaking through the party wall. On the first floor there is an office that has been extended into a third house. Above are three flats (Figure 4). If the buildings (the three former houses) were to be used as the units here, then the shop and the office would each be broken into two parts. What the NBD and 3DStock do in this case is to wrap up all three former houses into a self-contained unit (SCU), as shown by the red dashed line in Figure 4. Five premises are contained within the three buildings that form the SCU. The spatial relationship of premises to buildings in the SCU is not precisely specified. It may not be possible to match all energy meters precisely to their relevant users. But the totality of energy use can be related to the total floor area.

Figure 4

There are only two alternatives for the units of modelling in this kind of circumstance: either buildings as units or premises grouped into SCUs as units. Both involve approximations. However, the device of the SCU has the great merit of never breaking premises. If premises were split, e.g. by cutting the first-floor office in Figure 4 into three parts, how would the energy use attributable to the office be divided into three? How could an EPC for the same office be divided into three? How could a retrofit policy (e.g. a grant for a heat pump) be shared across the constituent parts of the divided office?

Such situations do not occur, obviously, with detached buildings and single occupiers. But they are surprisingly common in the non-domestic and mixed-use stock, especially along high streets in town centres, as will be shown below. The device of the SCU was first introduced by Simon Taylor in a model, built in the 2010s, of the non-domestic building stock of the city of Leicester (Taylor et al. 2014).

A non-domestic premises may occupy a single building, equating to a single SCU. Examples could include small primary schools or churches. At a larger scale, one premises, such as a secondary school, a large factory, a hospital or a university, might comprise many buildings on a shared site.³ It could often be the case that only one of these buildings contains the energy meters, carries the address and receives the mail. Other buildings would not have postal addresses. Gas and electricity supplies may be shared between many buildings, with pipes and private wires running between them. For these types of premises, 3DStock and the NBD define the relevant ‘campus SCUs’ by means of site boundaries with which all the buildings are lassoed and grouped together. Otherwise, many unaddressed buildings would escape the NBD, and floor areas would be radically underestimated.

As examples of the confusion of buildings with premises, the UK Green Building Council (UKGBC) has talked about there being 1.9 million non-domestic buildings in the UK in 2015 (UKGBC 2026). At that time there were 1.8 million premises in England and Wales (i.e. not including Scotland and Northern Ireland) listed by the VOA; premises are not buildings; and the coverage by the VOA is incomplete. The Climate Change Committee (CCC) quoted a figure of ‘over 1.9 million non-residential buildings in the UK’ in 2022: again, a mistaken equation of premises with buildings (CCC 2022).⁴ Within the NBD, there are 1.3 million SCUs containing 2.3 million non-domestic premises. Using those figures, a policy that treats premises as ‘buildings’ might greatly overstate the decarbonisation problem compared with one that deals with SCUs.

Researchers and decision-makers need to know about both buildings and premises, and the relationship between the two, because energy and adaptation measures are applied to buildings, but occupiers of premises make decisions about, and provide finance for, the installation of those measures.

2.2 MYTH 2: ‘THE NON-DOMESTIC STOCK CONSISTS LARGELY OF OFFICES’

In the NBD, activities in premises are classified by a system based in large part on the classifications used by the VOA, supplemented by further types employed in OS AddressBase data. The system is known as CaRBE, ‘Carbon Reduction in the Built Environment’. CaRBE distinguishes 365 activities at the lowest level, examples of which would be ‘bingo hall’, ‘lifeboat station’ and ‘garden centre’. These are aggregated for statistical purposes into 15 non-domestic classes, plus ‘Domestic’. It is possible to measure the magnitude of each class differently: by total floor area, by numbers of SCUs or numbers of premises (Table 1). (‘Domestic’ is not distinguished as a row in its own right here but represents around 30 million dwellings in Great Britain.) Within some SCUs there may be many premises, as, for example, in multi-tenant office buildings or warehouses divided into multiple units with different occupiers. This is reflected in the ratios of SCUs to premises shown in Table 1. (Other SCUs will equate to single buildings.)

Table 1 shows that there are both more SCUs and more premises in the Shop class than in the Office class; and that the total floor area of Shop is greater than that of Office. The total floor areas of the Factory and Warehouse classes are also greater than Office, although the numbers of premises are each lower, since floor areas in these classes are on average larger. The number of SCUs in Hospitality (hotels, pubs, restaurants, cafés, takeaways, etc.) is also greater than for Office, even though the total floor area of the class is smaller.

Clearly these numbers are dependent, to an extent, on exactly which lower-level activities are included in each class. The general picture is clear, however. Offices do not dominate the building stock and a decarbonisation policy designed to tackle just ‘Offices’ would only affect 13% of all non-domestic floor space, whilst Factory (20%), Warehouse (19%) and Shop (15%) all represent larger shares of total non-domestic floorspace. Might it be that those working on energy use in buildings in the past have chosen to study offices because they themselves worked in offices? Another explanation might be that offices seem to dominate the centres of towns and cities visually, while factories and warehouses are generally outside of central districts and out of sight. (But this would not account for shops and hospitality being largely ignored.)

Napoleon Bonaparte called Britain a nation of shopkeepers, but he did not mean it as an insult, rather an acknowledgement of the country’s commercial power. The NBD statistics confirm how large a floor area is still devoted to shops. Analysis of historical VOA data shows nevertheless that this area has been shrinking and becoming consolidated into fewer, larger shops in recent years, in part due to the competition from online retailing (VOA 2025). Meanwhile, the activity class that has been growing fastest is Warehouses, as the country pivots from a nation of shopkeepers to a nation of online shoppers.

2.3 MYTH 3: ‘NEW BUILDINGS ARE THE MOST IMPORTANT PART OF THE BUILDING STOCK’

This is the impression that might be gained from looking at illustrations in the architectural and building industry press. Magazines and websites tend to picture the building stock with photographs of shiny new offices and shopping centres. On the other hand, architectural historians and heritage organisations might want to challenge such a myth. The NBD replaces subjective impressions with a clear statistical picture. It has data on the ages of premises within ranges of dates, drawn from the VOA, EPCs and the OS National Geographic Database.

Table 2 shows the numbers and total floor areas of premises, expressed as percentages, falling in each age band. (Note that the bands are unequal in duration.) A total of 43% of premises by number date from before the Second World War, making up 24% of all floor area, showing that these older premises are generally smaller. The burst in construction that accompanied Britain’s recovery from the war is visible in the period 1940–75. Breaking down these figures by activity classes shows that Hospitality has most floor space dating from before 1919; while Health, Education and Warehouse have the most floorspace built recently, between 2007 and 2021.

The NBD also has data on buildings listed as being of special historical interest, and on which buildings fall within designated conservation areas. These show that 43% of all non-domestic premises are in a conservation area (31%) or are part of a listed building (4%), or both (8%). The activity classes that have the largest numbers of premises either listed or in conservation areas are Office and Hospitality, over 50% in both cases. Conservation areas prohibit or tightly control some permitted development rights for buildings (e.g. fitting external wall insulation, replacing windows or fitting solar panels). Any changes to a listed building require a Listed Building Consent (LBC). Together these limit the take-up of policies which encourage such measures. At the same time, newer buildings tend be more energy efficient and less likely to be within a conservation area or be listed. The result is that older buildings have a greater need for energy efficiency improvements, whilst at the same time being more likely to be listed or fall within a conservation area, creating a dilemma for policy teams.

2.4 MYTH 4: ‘THE BUILDING STOCK IS DIVIDED BETWEEN DOMESTIC AND NON-DOMESTIC’

Academic researchers and consultants working on energy use in buildings have tended to divide the field for convenience between domestic and non-domestic buildings. Within the UK government, energy use in housing has historically been the responsibility of a different department from that responsible for energy use in the non-domestic stock. Domestic energy use has rightly received more attention because the number of houses greatly exceeds the number of non-domestic premises. As mentioned, there are 30 million domestic addresses in Great Britain compared with 2.2 million non-domestic premises. The housing stock may also have been studied by preference because it is more homogeneous, and its patterns of energy use much less complex than the non-domestic stock. The consequence is that mixed-use buildings in which flats are combined with other uses have largely been overlooked.

A walk down any high street, however, shows how numerous these mixtures are: flats over shops, flats over cafés and restaurants, and flats over small ‘shop-like’ offices. These perceptions are quantified in the NBD. If the ‘pure’ unmixed housing stock is set aside, then in 27% of all remaining SCUs, some non-domestic activity is combined with one or more domestic addresses. This means that over a quarter of the non-domestic and mixed-use stock consists of domestic and non-domestic uses combined within the same buildings. There is the further possibility of different non-domestic activities being located together, as, for example, shops with offices. These account for a further 14%. Only 59% of these latter SCUs have an occupier or occupiers engaged in a single ‘pure’ non-domestic activity.

Figure 5 shows how these categories of SCUs—single-activity occupant; multiple-activity non-domestic occupants; mixed domestic and non-domestic occupants—vary in proportion between the activity classes. (The classes, indicated by code letters here, match the order given in Table 1.) There are more mixtures of flats with shops and of flats with hospitality premises than in other classes, as would be expected. Those in hospitality would include landlords’ accommodation in pubs. It might be a surprise that there are flats in the Factory class, but this is because it covers small workshops whose owners may ‘live above the shop’. Also, there can be caretakers’ and watchmen’s accommodation in factories.

Figure 5

There may be heat transfer between premises in mixed-use SCUs. A flat above a fast-food restaurant might benefit from the waste heat coming through the floor, but might have greater demands than otherwise for ventilation. These combinations of occupiers and their activities can introduce complexity in coordinating renovation or retrofit works in the shared buildings, as well as policy gaps when government policy is specifically only for non-domestic or domestic ‘buildings’.

2.5 MYTH 5: ‘OFFICES AND INDUSTRIAL PREMISES ARE LARGE’

The popular image of an ‘office building’ is a large, perhaps high-rise block in a city centre. Likewise, the word ‘factory’ can conjure up a sprawling manufacturing complex with an office building, storage sheds, maybe a canteen and sports facilities. If the basic unit is the premises and not the building, however, this means that many large office blocks are broken into numerous office suites or rooms with different occupants. The Office class in the NBD also covers the great variety of small office premises—lawyers, accountants, taxi offices, etc.—found on shopping streets. Some of these may consist of just one or two rooms.

Meanwhile, the Factory class covers workshops, whose median floor area in the NBD is 136 m², and vehicle repair workshops, whose median area is 177 m². In both classes, Office and Factory, the overall distribution of size (measured by floor area) is heavily skewed, with relatively large numbers of small premises and long tails of bigger premises. The mean area of all premises in the Office class is 246 m², and in the Factory class is 609 m².

This skewing of size, however, has a positive implication for policies to cut emissions. A small number of very large buildings in each class accounts for a significant fraction of the total floor area. Table 3 shows the percentile share of the top 100, 200, 500 and 1000 premises in the largest five CaRBE classes, demonstrating how for Factory, Hospitality, Office and Warehouse, the top 1000 premises account for over 15% of total floorspace. Notice how the top 1000 Office premises account for 0.21% of all Office premises yet represent nearly 19% of total floorspace.

The logistical and political task is thus simplified by the possibility of dealing with a small number of owners, who are in a better position to mobilise financial and organisational resources than the many owners of small premises, and who can make the biggest contributions to cutting emissions.

2.6 MYTH 6: THE BUILDING STOCK CAN BE CLASSIFIED FOR ENERGY ANALYSIS BY THE STANDARD INDUSTRIAL CLASSIFICATION (SIC)

The SIC is a set of codes, first introduced in the United States in the 1930s, for classifying economic activities (ONS 2016). It has become widely used by government and commerce in many countries, including the UK, for compiling economic and employment statistics. The latest UK revision was in 2007. The SIC consists of four-digit numerical codes, qualified further in some cases by a fifth digit. Some random examples of categories are ‘deep coal mine’, ‘family welfare association’ and ‘feather duster (manufacture)’.

The SIC is used in the UK at a highly aggregated level for publishing national energy statistics, in particular the annual Digest of United Kingdom Energy Statistics (DUKES). SIC categories are also used in the government’s UK TIMES model of the national energy system. Why then is the SIC not used for the NBD?

There are instances where CaRBE and SIC coincide. But the overriding issue is that CaRBE describes the activities undertaken in premises and buildings, while the SIC describes the economic sectors within which those activities fall. A major chain of supermarkets, for example, would own the shops themselves, as well as storage warehouses and a head office building. In the SIC, these would all be classified as ‘Food (general) (retail)’ (47110). In CaRBE, by contrast, they would be classified variously as ‘supermarket/hypermarket/superstore’, ‘warehouse’ or perhaps ‘cold store’, and ‘office’. Similarly, the head office of a petrol company would be classified by the SIC as ‘Petroleum refining (manufacture)’ (19201) and ‘Petroleum products distribution (wholesale)’ (46711); while in CaRBE it would be ‘office’. Clearly CaRBE classifications, since they relate to premises and buildings, not companies, are in general much better suited to energy analysis of the building stock.

A major problem arises with warehouses which, as has been shown, make up a large fraction of the non-domestic stock. DUKES classifies end users of energy by nine high-level SIC groupings: fuel producers; iron and steel; other industry; transport; agriculture; commercial; public administration; other services; and domestic. In the UK TIMES model, these are further compressed into just five groups. However, warehouse buildings are found in practice in the industrial sector, in the commercial sector (retail warehouses), as well as in other sectors including government, health, the military, etc. There is thus no indication in DUKES or UK TIMES of how warehouses as an activity class, defined in building-related terms, are divided between the respective SIC groupings. Work is in progress to bring together the CaRBE classification and 3DStock with UK TIMES, which should allow for more detailed, lower-level modelling of energy use in the building stock.

On the other hand, there is one area in which the SIC could in theory be a useful adjunct to the NBD and CaRBE system. The CaRBE Factory class, as mentioned, covers many premises by the simple description ‘factory’ or ‘workshop’, without further detail. If SIC classifications were added in such cases, these would indicate the types of commodities being produced and could give further information about likely equipment and energy use. In a similar way, many small shops are classified as such in CaRBE without further qualification. Here the SIC could help in distinguishing the types of goods being sold. Efforts have been made in the past to attach SIC classifications to premises in models, but so far without success. The difficulty is that SIC codes are assigned in government and commercial databases to companies and institutions, not separately to the individual sites and buildings they occupy.

2.7 MYTH 7: THE MYTH OF THE REPRESENTATIVE ENERGY BENCHMARK

Government bodies and organisations connected with the building industry have published energy benchmarks to provide designers with indications of average levels of energy use in buildings of different types, and to allow them to compare these with best practice. In Britain, the leading role has been played by the Chartered Institution of Building Services Engineers (CIBSE) through its TM46 report on Energy Benchmarks (2008) (CIBSE 2008). More recently, CIBSE has been developing an Energy Benchmarking Dashboard, available online (CIBSE 2025). In both cases, the benchmarks are for buildings classified by activities—in a few cases qualified by size and age of building. The 2008 edition of TM46 listed 196 activities, grouped into 29 categories for which benchmarks were provided. This coverage was subsequently extended with an analysis of Display Energy Certificates (DECs) for public buildings in England and Wales, in which actual annual energy consumption is recorded (Hong & Steadman 2013). The online tool gives 126 distinct benchmarks for activity types, separated into electricity and fossil fuels, with figures for ‘typical practice’ and ‘good practice’ in kWh/m²/yr.

These guidelines have served useful purposes. It is reasonable to suppose that there is some consistency in energy use within certain large groups in which activities are relatively homogeneous, such as houses, offices and schools. But benchmarks—with the exception of those for houses—have, for lack of available data, tended to ignore differences in building size, form, construction, services and equipment, all of which vary widely in the non-domestic stock, and which can have large consequences for energy intensities. (There was a debate within the CIBSE Benchmarks Committee as to whether a separate higher benchmark should be set for offices with air-conditioning. It was decided not to do this to avoid creating perverse incentives.) NBD now changes this situation.

The main problem with benchmarking schemes to date is that they have been based on information from small numbers of buildings, obtained opportunistically, with little knowledge of whether these samples were representative of the complete stock of each type. The sources for the data behind TM46, and the size of the samples measured, have never been made public. If inaccurate benchmarks are based on limited samples, potential users can lose confidence. This happened with the oddly named Julie’s Bicycle, a coalition of arts and cultural organisations campaigning for climate action (Julie’s Bicycle 2025). Julie’s Bicycle found that energy benchmarks in TM46 for cinemas and theatres did not correspond to their experience and began to collect their own data from meter readings. Other organisations that were not so sceptical might have been in danger of making decisions on similarly misleading benchmarks.

There is little purpose in trying to make a systematic comparison between historical benchmark values and values from the NBD because of the great differences in activity classification, the relative sizes of the samples and the passage of time. Solely for illustration, however, a few benchmarks (medians for electricity and fossil fuels combined, kWh/m²/yr) from three sources can be compared: surveys made by a team from Sheffield Hallam University (SHU) in the 1990s (Mortimer et al. 1999), TM46 and the NBD (Table 4).

The values for offices and hotels in the NBD are plausibly lower because of the inclusion of large numbers of small premises in the case of offices, and many guesthouses and bed-and-breakfasts in hotels.

The problem is that there has not been any publicly available national building census against which the representativeness of benchmarks could be measured.⁵ Such a census now exists, however, in the form of the NBD. The number of activities distinguished in the NBD by CaRBE classification, as mentioned, is 365, and much more detailed than previous benchmarking schemes. Because of their basis in the classifications developed over many years by the VOA and OS, it is certain that their coverage is complete. The matching of energy meters to premises in the NBD is not universal, for various technical reasons. Overall, the rate of matching electricity meters to premises is 48%, and of gas meters to premises is 22%.

The matching is not uniform and some CaRBE classes match better than others. For example, the match rate for electricity meters for Shop is around 72%, whereas it is as low as 20% for Utilities. Also note that some premises are off the gas grid, so would not have gas meters. Taking this into account suggests that the reported 22% match rate for gas should be more like a 25% match rate when off-gas postcodes are excluded. The total number of premises in the NBD is 2.2 million. This means that more than 1 million premises have electricity meters matched, and just under 0.5 million have gas meters matched. The NBD thus provides energy benchmarks from the largest sample that can in practice be assembled. The complete distributions of values can be plotted, with their means and medians. These values have been published by DESNZ (2026) in a form that is not disclosive of the consumption of individual premises including medians, interquartile ranges (IQRs), whisker plots and counts.

Because premises are related to buildings in every case in the NBD, it will be possible in principle to go further and analyse how patterns of energy consumption within some given activity class vary according to building characteristics such as size, height, age and construction, as well as by their servicing systems. Within offices, for example, variations can be analysed by height, or by whether buildings have air-conditioning, or have curtain walls, all of which can significantly affect consumption.

Issues will remain, however, arising out of the heterogeneity of some activity classes, and from the very concept of the benchmark. Two places where the CaRBE classification is weakest, due to the sources on which it draws, are the Shop and Factory classes, as already discussed. A major study of energy use in non-domestic buildings made in the 1990s by a team from Sheffield Hallam University (SHU) showed how the type of goods sold in shops can have a large impact on energy use, ranging from very low consumption in flower and fish shops, kept open to the exterior, at one extreme, to shops using large amounts of refrigeration and even, like bakers, manufacturing goods on-site, at the other extreme (Ashley et al. 1997). This wide range is reflected in the aggregate statistics in the NBD for total energy use by small shops—grouped together with similar sized retail premises such as bank branches and showrooms which have IQRs for total energy of 138 and 99 kWh/m²/yr, respectively, when analysed at the individual CaRBE activity level, but have an IQR of 248 kWh/m²/yr when treated as Shops as an entire CaRBE class.

A similar situation exists with types in the Factory class, where the manufacture of some types of products is separated out, but there remain large numbers of ‘factories’ and ‘workshops’ that are not further classified. In many of these the energy demand will be dominated by the equipment used.

Benchmarks calculated for large generic activity classes can conceal the fact that within these are some especially high energy users deserving separate consideration. These can be identified in the NBD: data centres, fast-food takeaways, launderettes, leisure centres with heated swimming pools and crematoria. The converse is also true, for example, with Workshops, which represent 76% of the Factory class yet have a median electrical energy use intensity of less than 40 kWh/m²/yr. This means that a decarbonisation policy targeting factories, but ignoring workshops might overstate the magnitude of the decarbonisation potential.

Benchmarks are typically normalised to floorspace (e.g. 110 kWh/m²/yr) and as such they represent an average for the whole premises. Within some premises there may be large parts that are not heated or cooled. A recent example here was the Future Buildings Standard consultation (MHCLG 2023), which focused on setting new, more stringent energy efficiency and carbon standards for new non-domestic buildings. For warehouses this was set with the assumption that 50% of the floorspace would be ‘conditioned’ (heated or cooled) based upon a ‘consultant-derived assumption made at the time due to a lack of robust empirical evidence’ (MHCLG 2026a: 51). Upon consultation on the changes to Part 6 and Part L (conservation of fuel and power), Part F (ventilation) and Part O (overheating) (MHCLG 2026a), this figure was revised and published (MHCLG 2026b) based upon evidence supplied by the NBD.

Using NBD and VOA data, this question was investigated at the level of spaces and rooms within warehouse premises. The results (Figure 6) show that for certain activities within the Warehouse class, in particular cold stores, the extent of conditioned floor space was more than 75% on average, as might be expected; but for standard warehouses, large distribution warehouses and wholesale warehouses this figure was less than 10%. The overall figure for the Warehouse class as a whole was closer to 13% on average. The 50% assumption would have implied that 86 million m² of floor space would need decarbonising, while the lower figure of 10% would mean only 22 million m². A revised 15–20% range was agreed (MHCLG 2026a: 51, 84), providing a real-world illustration of how ‘consultant derived assumptions’ can result in unrealistic predictions. Non-domestic buildings are far more heterogeneous than houses. ‘The devil is in the detail.’

Figure 6

There is then the major phenomenon of mixed-use buildings, where it is unclear what an energy benchmark means or how it should be applied. Should there be special benchmarks for specific mixtures, as, for example, for shops with flats above, or offices with shops? The numbers of alternatives could be large, especially if the proportions of floor area devoted to each use were taken into account. Or should a benchmark for each ‘pure’ activity be applied to the premises to which it applies, in proportion to the fraction of floor area occupied? This is the approach recommended by CIBSE (2008) in TM46. However, it assumes in effect that there is no thermal interaction between premises; and that typical levels of consumption for a given activity are similar for freestanding premises as for premises within mixed-use buildings, which is questionable. Light may be thrown on these issues by further analysis of the NBD.

2.8 MYTH 8: ‘ENERGY PERFORMANCE CERTIFICATES (EPCS) AND DISPLAY ENERGY CERTIFICATES (DECS) CAN PROVIDE A ROBUST, UNBIASED SAMPLE OF THE NON-DOMESTIC BUILDING STOCK’

Non-domestic EPCs are mandatory in England and Wales when buildings/premises are built, sold or rented. They rate theoretical energy efficiency of a building from A (most efficient) to G (least efficient), and provide recommendations for improvement. DECs are currently required for buildings over 250 m² in size that are frequently visited by the public, e.g. hospitals, schools and leisure centres. They are also rated from A to G, but this rating is based on actual energy consumption.

Over time, EPCs and DECs have been lodged in the Energy Performance of Buildings Register (EPBR), which now contains around 1.5 million non-domestic EPCs and just over 0.5 million DECs. The Register is held by the Ministry of Housing, Communities and Local Government (MHCLG).⁶ Using these totals, it might be easy to believe that the combined 2.06 million certificates would provide an excellent sample of the 2.07 million non-domestic premises in England and Wales. However, this is a false assumption for two main reasons.

These concern the cross-referencing of EPCs to premises, and the fact that many duplicate certificates exist in the Register. Ony 64% of EPCs and 77% of DECs have unique property reference numbers (UPRNs) that can link them easily to addresses. As for duplication, analysis shows that around 7% of EPCs are repeated in the Register, while a massive 77% of DECs are multiple records for the same premises. Thus, the true numbers of unique non-domestic certificates with a UPRN are likely to be nearer 800,000 EPCs and 86,000 DECs.

What is more, EPCs are only collected for specific regulatory reasons (‘triggers’) rather than as a representative survey of the non-domestic stock. The MHCLG itself notes this fact:

These ‘triggers’ for certificates mean they will overrepresent recently sold, or rented, or new/refurbished buildings, and underrepresent older premises and buildings with long-term owners or tenants. Some EPCs are triggered when improvements are made to a property, either through the provision of a government grant or through refurbishment. This can bias the ‘EPC stock’ to appear newer and more energy-efficient than the complete population of premises.

There are other ways in which EPCs are unrepresentative of the stock. Some non-domestic activities, oddly, take place in domestic addresses with domestic EPCs. One such example is provided by self-catering holiday homes, which count as non-domestic since they are businesses, and which represent a large percentage of Hospitality premises. By taking only non-domestic EPCs, the majority of these premises would be excluded from analysis of the non-domestic stock. Some premises will have a certificate but can opt out of making this publicly accessible. Other certificates are excluded from the EPBR on the grounds of national security. This can, for example, result in an underrepresentation of Ministry of Defence (MoD) premises. Geographical bias can occur by the very fact that certain parts of the country see higher turnovers of property. This can mean that wealthier regions (e.g. Greater London, South East England) are overrepresented, whilst rural areas with low value, non-domestic property are underrepresented.

Some premises in certain activity classes change hands very infrequently. Factories are a case where turnover is infrequent, and premises are sold or have a change of tenancy far less often than Shops or Hospitality. This is reflected in the 24% of premises in the NBD which have an EPC or a DEC in the Factory class, compared with 51% in Shops and 55% in Hospitality. There are yet further issues relating to inconsistent classifications of activities in non-domestic EPCs.

Turning to DECs: while they are required for all publicly accessed properties sized over 250 m² (since 2015), before this date the threshold was 500 m² in 2013 and 1000 m² in 2008. This can mean that smaller premises are still underrepresented and that DECs are biased towards larger public sector buildings. Also, compliance in registering DECs is low for certain building types.

The combination of these factors means that while it is possible to draw what appears to be a very large sample of non-domestic premises from the EPC and DEC records held in the EPBR, these are administrative datasets not population surveys. They are skewed towards newer, larger premises. The number of workable records in the EPBR is far lower than the total number of records (and a failure to appreciate this will result in double, or multiple, counting). Treating this dataset as representative would create systematic optimism bias: it would give the impression that the non-domestic building stock is newer and more energy-efficient than it really is, and that premises and buildings are larger in terms of floorspace than they really are. EPCs, like previous benchmarking schemes, give an unrepresentative picture of the stock.

Non-domestic EPCs could in theory be based in the future on a combination of actual energy-use data (as with DECs) with other attributes taken from an augmented NBD. This would offer a more consistent assessment of premises/buildings that could apply internationally, and allow international investors, including banks, to compare more effectively between investment decisions in different countries.

2.9 MYTH 9: ‘THE MYTH OF THE REPRESENTATIVE ARCHETYPAL BUILDING FOR URBAN ENERGY MODELLING’

The NBD is a snapshot of the stock as it existed in April 2023.⁷ (It is to be updated regularly.) To predict future trends and the effects of applying new measures and policies, it has been general practice to use building energy simulation. Because of the heavy computational demand of these models, it is not practical to simulate every building in a region or city individually—even if adequately descriptive data on all buildings were available. Recourse has been made in these circumstances to a reduced set of archetypal buildings, supposedly representative of the given stock, each of which is modelled separately. Results are then grossed up to the sector or region by multiplying by the relevant number of buildings.

This use of archetypes for non-domestic buildings faces the same difficulties as those discussed in relation to benchmarks. How are the archetypes derived, and how is it possible to know whether they are representative, without a census of all buildings? It seems that archetypes for stock models have on occasion been drawn from the imagination or from professionals’ subjective experience, subject to all the myths listed previously.

Reinhart & Davila (2016) make an international review of the field of urban and regional building stock modelling. They report models using between 10 and 150 archetypes in total. Comparing model predictions with total actual consumption, Reinhart and Davila found error ranges of 12–55% for regional stock models and of 5–99% for urban models. Figure 7 shows two archetypal office buildings, ‘large’ and ‘small’, used by Ding & Zhou (2020) to cover the entire office stock of the Chinese city of Wuhan. This is an extreme example, but it illustrates how archetypes can fail to address the variety of size, construction and other characteristics in the real stock. The fact that archetypes may not be representative is likely to be just one cause of the inaccuracies reported by Reinhart & Davila (2016). Also, well-known uncertainties arise in the simulation of complex buildings, in particular the modelling of infiltration rates and occupant behaviour. Large errors in the predictions made by these models could lead to misdirected and ineffective policies at the urban and regional scales.

Figure 7

It is instructive to compare the situation for the domestic stock. In Britain, the principal instrument for energy policy is the National Household Model (NHM) (BEIS 2017). This uses the English Housing Survey (EHS) to provide a detailed census. The NHM takes 11,000 dwellings from the EHS to represent the total of some 30 million houses and flats in the UK. The activity is, of course, the same in all cases. These are not artificial archetypes but are actual houses and flats selected to be representative of a series of other characteristics, including built form (flat, terrace, semi-detached, detached, etc.), size, geographical region, age band, heating system type, and wall and roof construction.

Given that there are 2.2 million non-domestic premises in Great Britain, how many archetypes or actual buildings would be needed to represent the non-domestic stock to a level of detail comparable with the NHM—given the much greater heterogeneity not only in activity but also in all other characteristics, and not forgetting that while every house is a separate building/SCU, non-domestic premises can be combined in complex arrangements in mixed-use SCUs?

The NBD provides opportunities for a completely new approach to the definition of archetypes. Instead of being imagined ad hoc, they can be generated by statistical analysis of all premises in the relevant sectors of the stock. Alternatively, there is potential for using cluster analysis, where statistical analyses groups together premises across the whole stock that are similar in terms of energy use, built form, construction type or whatever characteristics are relevant. This can guarantee their representativeness. Simulation models can be calibrated against actual consumption, since this is known for the great majority of premises.