Publication - Research and analysis

Developing regulation of energy efficiency of private sector housing (REEPS): modelling improvements to the target stock - Main Research Report

Published: 5 Nov 2015
Part of:
Research
ISBN:
9781785447730

This report describes how the least energy efficient dwellings in the private sector were identified and how their ratings could be improved by a range of improvement measures. Modelling was used to ascertain the least cost way of reaching different standards, with findings presented on capital costs, fuel cost savings, carbon and energy reductions.

260 page PDF

7.2 MB

260 page PDF

7.2 MB

Contents
Developing regulation of energy efficiency of private sector housing (REEPS): modelling improvements to the target stock - Main Research Report
3 Developing A Typology Of The Private Housing Stock In Epc Bands EFG

260 page PDF

7.2 MB

3 Developing A Typology Of The Private Housing Stock In Epc Bands EFG

3.1 In this section, we describe the underlying principles and methods for creating the typology of the target stock. We then outline the typology groupings and how these were represented by archetype properties. Overall, the methods outlined resulted in 355 dwelling archetypes for modelling improvements.

Rationale underpinning the typology of the target stock

3.2 Fundamentally, developing a typology involved splitting the private sector housing stock in bands E, F and G into groupings of similar types of properties. Central to the creation of the typology groups was the need to minimise variation within groups.

3.3 One dwelling - an archetype - then had to be chosen to represent each typology grouping. Archetypes therefore had to represent all the dwellings within a typology group and needed to be an average example, typical of the typology group.

3.4 Given that the policy options had not been set at the time the typology groups were identified, the approach taken needed to allow a variety of potential options to be examined. The typology needed to cover as much as possible of the target stock - private sector dwellings in EPC bands E, F and G -and be designed to assess both where potential improvements could be made and the likely impact of improvements.

3.5 The assessment (and prioritisation) of potential improvement measures for each archetype followed a logical order:

  • Is the measure technically feasible?[18]
  • If yes, would it lead to gains in energy efficiency?
  • If yes, how costly would it be to implement the measure?

3.6 Therefore, similarity between dwellings within a typology grouping had to be defined primarily in terms of characteristics that determine the technical feasibility of improvement measures and likely gains in energy efficiency, rather than, for example, characteristics that are more closely related to costs.

3.7 While dwellings with similar construction, insulation and heating characteristics were likely to have similar energy efficiency ratings regardless of floor size (SAP was designed to be insensitive to floor area), the reverse does not hold. Dwellings with similar energy ratings will vary considerably in terms of their form.

3.8 There are two potential approaches to constructing typologies: a top down approach, whereby the start-point is 100% of the target stock and this is then continually sub-divided until reaching suitably homogenous groups; or a bottom up approach, where key segmentation variables are defined, all combinations of which are created, and then groupings are added together or split further.

3.9 Because of the importance of key build characteristics, a bottom-up two-stage approach was taken. The initial stage involved segmentation based on four key dwelling characteristics. In the second stage, relatively large groupings were then further sub-divided (by EPC band and in some instances size of dwelling) while groupings that accounted for a very small proportion of the target stock were combined back into more prevalent groupings.

3.10 The initial stage was to segment the dwelling stock on four key dwelling characteristics: dwelling type, dwelling age, wall type, and main heating type. These are key variables in RdSAP and their inclusion helped ensure that the typology groups were homogenous in relation to the technical feasibility of measures and likely gains in energy efficiency.

  • Dwelling type: At one level built form is not important to the energy analysis of the dwelling stock, as the modelling was undertaken via a full SAP program (that does not use the built form to determine any of the dimensional details such as surface areas and heat loss characteristics of walls, floors and roofs which happens in RdSAP). However, built form is the way most people categorise and visualise the dwelling stock. It also acts as a good proxy for surface areas and heat loss characteristics of walls, floors and roofs. The distinction between houses and flats was also important in relation to assessing the cost of works.
  • Dwelling Age: Building standards have tended to improve over time. Age is an essential variable within the RdSAP process because, when coupled with the wall construction or the floor construction, it determines the assignation of the respective default U-value[19]. It is also used to assign a U-value when the insulation levels of the dwelling are not known as well as helping to determine the default window area in houses and flats.
  • Wall type: Wall construction is important factor in assigning the default U-value to a wall type, which then is a major determinant in the calculation of heat loss from the dwelling. Walls are usually the largest heat loss surface area in a dwelling. The wall construction is also a significant determinant in whether lower cost energy efficiency improvements (such as cavity wall insulation) can be undertaken on the property, or whether more expensive internal or external wall insulation is needed.
  • Main heating system: The single most significant factor in determining the energy rating of a dwelling, its carbon emissions, and level of fuel bills is the combination of fuel and heating system efficiency. Highly efficient heating such as direct acting electric heating is expensive to run, and has a high CO2e co-efficient attached to it. Similarly, there is a significant difference in fuel bills for similarly efficient mains gas and LPG boilers because of the significant differences in the associated cost of fuel. Some cheaper to use fuels have a higher carbon co-efficient associated with them (such as solid fuel) compared to wood or biomass[20].

3.11 Some sub-categories of these four key characteristics were combined, where possible, before the initial typology was created. This was only undertaken on a limited number of categories to ensure that there was a low risk of grouping together dwellings with very different characteristics.

  • Dwelling type: End-terraced houses were grouped with semi-detached houses, while enclosed end terraced houses were grouped with mid-terraced houses (both on the basis of number of party walls) to give 7 groups overall[21].
  • NHER ages were combined into 6 bands based on the similarity of U-values amongst wall constructions. Further collapsing would have meant different default U-values for wall construction in RdSAP[22] in the same banding.
  • Wall type: Cob (earth) wall type dwellings (that made up less than 0.5% of the target sock) were grouped with Solid Brick to give 6 bands[23].
  • Main heating type was reduced into 13 bands by collapsing some of the rarer forms of heating that share common characteristics, e.g. all warm air systems were combined together, as were heat pumps regardless of whether they were using on or off peak electricity[24].

3.12 These four criteria were used as the basis for establishing the typology. When combined, 7 dwelling types, 6 NHER age bands, 6 wall type categories and 13 main heating forms give a theoretical total of 3,276 possible combinations. However, only 325 combinations existed in the SHCS data.

3.13 These 325 combinations differed considerably in their prevalence within the target stock:

  • High prevalence: 13 combinations each accounted for more than 1.5% of the target stock (over approximately 6,450 dwellings each) and 38% overall (153,500 dwellings).
  • Medium prevalence: 29 combinations each covered between 0.5% and 1.5% of the target stock (between approximately 2,000 and 6,000 dwellings) and 26% overall (102,200 dwellings).
  • Low prevalence: 126 combinations each covered between 0.1% and 0.5% of the target stock (between approximately 400 to 1,950 dwellings) and 27% overall (110,250 dwellings).
  • Very low prevalence: 157 combinations covered less than 0.1% of the target stock each (less than 400 dwellings) and 9% overall (34,600).

3.14 The high, medium and low prevalence combinations were further split while the very low prevalence combinations were collapsed in with the most similar grouping among the higher prevalence groupings where possible.

3.15 The high prevalence combinations were further split by up to four SAP 2005 score bandings - upper EPC band E (SAP 47-54), lower band E (SAP 39-46)[25], F (21-38) and G (20 and below) - and by 2 floor size bands[26]. Overall, this gave 76 sub-groupings among the 9 high prevalence combinations (as shown in Table 3.1).

3.16 To illustrate, the most prevalent dwelling combination, as shown at the top of Table 3.1, was pre-1919 detached houses with solid granite walls and an oil-boiler based heating system. These accounted for 5% of the target stock (19,000 dwellings). These dwellings were further sub-divided into 6 sub-groups - two different floor sizes[27] within three EPC groupings, upper E, lower E, and band F - each to be included in the final typology and each with a different archetype. Note that none of the cases in the data for this combination of building characteristics were in EPC band G.

Table 3.1: Summary of the high prevalence combinations and proposed sub-groups

Initial combinations Further split Total sub-groups Dwellings
(%age of target stock)
Sample size %age in Upper E
(47-54)
%age in lower E
(39-46)
%age
in F
%age in G
Pre-1919-Detached-Granite-Oil boiler x3 EPC bands, x2 Size 6 19,900
5%
59 20% 44% 36%  
Pre-1919-Detached--Sandstone-Oil boiler X4 EPC bands, x2 Size 7[28] 17,800
4.4%
104 28% 38% 33% 1%
1919-1964-Detached-Cavity Brick-Mains Gas boiler x3 EPC bands, x2 Size 6 17,700
4.4%
71 75% 20% 5%  
1919-1964-Semi/End Terraced-Cavity Brick-Mains Gas boiler x2 EPC bands, x2 Size 4 14,850
3.7%
64 96% 4%    
Pre-1919-Semi/End Terraced--Sandstone-Mains Gas boiler x3 EPC bands, x2 Size 6 13,200
3.3%
48 72% 23% 5%  
1965-1975-Detached-Cavity Brick-Mains Gas boiler x3 EPC bands, x2 Size 6 12,300
3.1%
53 77% 19% 5%  
Pre-1919-Detached--Sandstone-Mains Gas boiler x3 EPC bands, x2 Size 6 11,550
2.9%
44 39% 43% 17%  
Pre-1919-Tenement-Sandstone-Electric peak room heater x4 EPC bands, x2 Size 8 8,950
2.2%
26 12% 16% 48% 25%
Pre-1919-Tenement-Sandstone-Mains Gas boiler x2 EPC bands, x2 Size 4 8,050
2.0%
22 65% 35%    
Pre-1919-Detached--Granite-Mains Gas boiler x3 EPC bands, x2 Size 6 7,950
2.0%
29 59% 33% 7%  
1965-1975-Semi/End Terraced-Cavity Brick-Mains Gas boiler x2 EPC bands, x2 Size 4 7,750
1.9%
29 93% 7%    
1919-1964-Semi/End Terraced-Cavity Brick-Off Peak Electric storage heating x3 EPC bands, x2 Size 6 7,050
1.8%
29 47% 32% 21%  
Pre-1919-Detached--Granite-Off Peak Electric storage heating x4 EPC bands, x2 Size 8 6,450
1.6%
19 3% 23% 58% 16%

3.17 The 29 medium prevalence combinations were further split into up to 4 separate typology groups based on the EPC bandings (with EPC band E split again into upper and lower). This resulted in 79 typology groupings. These are detailed in Table 3.2.

Table 3.2: Summary of the medium prevalence combinations and proposed sub-groups

Initial combination Further split Sub-groups Dwellings
(%age of target stock)
Sample size %age in Upper E
(47-54)
%age in lower E
(39-46)
%
in F
% in G
1965-1975-Semi/End Terraced-Cavity Brick-Off Peak Electric storage heating x3 EPC bands 3 5700
1.4%
29 56% 35% 9%  
Pre-1919-Semi/End Terraced-Solid brick/Cob-Mains Gas boiler x2 EPC bands 2 5450
1.4%
25 73% 27%    
Pre-1919-Semi/End Terraced-Granite-Mains Gas boiler x3 EPC bands 3 5450
1.4%
21 62% 28% 10%  
1919-1964-Semi/End Terraced-Solid brick/Cob-Mains Gas boiler x2 EPC bands 2 5150
1.3%
18 87% 13%    
Pre-1919-Flat from converted house-Sandstone-Mains Gas boiler x3 EPC bands 3 5100
1.3%
20 43% 45% 12%  
1919-1964-Detached--Cavity Brick-Oil boiler x3 EPC bands 3 5000
1.2%
46 51% 41% 8%  
1965-1975-Detached-Cavity Brick-Oil boiler x3 EPC bands 3 4950
1.2%
36 44% 52% 4%  
Pre-1919-Tenement-Sandstone-Off Peak Electric storage heating x3 EPC bands 3 4450
1.1%
14 56% 16% 28%  
Pre-1919-Semi/End Terraced-Sandstone-Oil boiler x3 EPC bands 3 4400
1.1%
25 59% 23% 19%  
Pre-1919-Semi/End Terraced-Sandstone-Off Peak Electric storage x4 EPC bands 4 4250
1.1%
24 23% 50% 23% 5%
Pre-1919-Detached-Solid brick/Cob-Mains Gas boiler x4 EPC bands 4 3950
1.0%
14 24% 43% 25% 9%
1919-1964-Tenement-Cavity Brick-Mains Gas boiler x2 EPC bands 2 3850
1.0%
12 67% 33%    
Pre-1919-Semi/End Terraced-Granite-Off Peak Electric storage heating x3 EPC bands 3 3550
0.9%
9 13% 65% 22%  
Pre-1919-Detached--Sandstone-Off Peak Electric storage heating x4 EPC bands 4 3400
0.9%
36 10% 22% 62% 6%
1919-1964-Semi/End Terraced-Cavity Brick-Oil boiler x2 EPC bands 2 3200
0.8%
15 89% 11%    
1919-1964-Detached-Solid brick/Cob-Mains Gas boiler x3 EPC bands 3 3150
0.8%
15 48% 42% 10%  
Pre-1919-Tenement-Sandstone-Means gas room heater x3 EPC bands 3 2650
0.7%
8 60% 21% 19%  
Pre-1919-Detached-Sandstone-LPG boiler x2 EPC bands 2 2600
0.6%
11     25% 75%
1976-1983-Detached-Cavity Brick-Oil boiler x2 EPC bands 2 2600
0.6%
17 76% 24%    
Pre-1919-Tenement-Granite-Off Peak Electric storage heating x3 EPC bands 3 2550
0.6%
9 50% 48% 1%  
Pre-1919-Semi/End Terraced-Granite-Oil boiler x3 EPC bands 3 2550
0.6%
10 48% 2% 50%  
1919-1964-Detached-Cavity Brick-Off Peak Electric storage heating x3 EPC bands 3 2500
0.6%
17 26% 44% 30%  
1965-1975-Detached-Cavity Brick-Off Peak Electric storage heating x4 EPC bands 4 2450
0.6%
21 15% 31% 42% 12%
1919-1964-Tenement--Cavity Brick-Off Peak Electric storage heating x3 EPC bands 3 2400
0.6%
7 74% 11% 15%  
1919-1964-Semi/End Terraced-System-Mains Gas boiler   1 2400
0.6%
12 100%      
1919-1964-Detached-Granite-Mains Gas boiler x2 EPC bands 2 2350
0.6%
7 77% 23%    
1919-1964-4-in-a-block-Cavity Brick-Mains Gas boiler x2 EPC bands 2 2100
0.5%
8 83% 17%    
1919-1964-Mid-ter/Ter with passage-Cavity Brick-Off Peak Elec storage x2 EPC bands 2 2050
0.5%
8 66% 34%    
1919-1964-Semi/End Terraced-Timber-Mains Gas boiler x2 EPC bands 2 2000
0.5%
6 68% 32%    

3.18 The 126 low prevalence combinations, that accounted for between 0.1% and 0.5% of the target stock (400 to 1,950 dwellings), were split into a maximum of three sub-groupings based on the EPC bands. (These are detailed in Appendix 4.) This was a similar approach to the medium prevalence combinations but did not make a distinction between dwelling in upper and lower halves of band E. Note, however, that many of these combinations only have dwellings in one EPC band and overall, this resulted in 188 archetypes.

Collapsing the very low prevalence typology groups into higher prevalence groups

3.19 The 157 very low prevalence combinations are detailed in Appendix 5. Given that each of these only account for a very small proportion of target stock - each less than 0.1% of the target stock (500 dwellings) and 9% overall - these were not modelled separately. Instead, they were collapsed backed into similar archetypes among the more prevalent groupings.

3.20 Central to the approach used to segment the target stock into archetype groups was the minimisation of variation within groups. This was also central to the approach taken to grouping the very low prevalent typology groups into similar archetypes among the more prevalent groupings.

3.21 A key consideration was the relative importance of the four building characteristics used to create the initial groupings - age and type of dwelling, wall type and heating type - plus EPC in this matching process. As heating type and wall type were more closely related to the feasibility of different technical improvements than dwelling type and dwelling age, and more important in determining energy efficiency ratings, these factors were prioritised.

3.22 A three-stage approach was undertaken:

  • Stage 1: Relax house type (to house/flat) and age band of dwelling criteria. Match with a more prevalent typology group if heating type AND wall type AND EPC banding AND broad dwelling type can be matched. If possible, also match on either age band of dwelling OR dwelling type. Overall, 81 of the 157 very low prevalence groups were matched in this way. They accounted for 5% of the target stock overall.
  • Stage 2: Match with a more prevalent typology group if they share EPC banding and EITHER wall type OR fuel type. If possible, also match on broad dwelling type (house/flat) and age band of dwelling where possible. Overall, 43 of the very low prevalence groups (3% of the REEPS target stock) were matched in this way.
  • Stage 3: The remaining 33 very low prevalence groups accounted for less than 2% of the target stock. These were individually inspected. Overall, 9 of these were collapsed into more prevalent archetypes while an additional 12 archetypes were modelled to cover the remaining 24 very low prevalence groups.

3.23 In total, 355 typology groupings were created. On average, each typology group represented around 1,100 dwellings (0.3% of the target stock), ranging from 30 dwellings (< 0.01% of the target stock) to 7,300 dwellings (1.5% of the target stock). Appendix 6 details the full list of typology groupings modelled against the initial combination of typology factors.

Archetype selection from within typology groups

3.24 The archetype chosen to represent each typology group had to be typical, and average with regard to any variation within the groupings.

3.25 As heating type, wall type, age and dwelling type were central to the typology creation, these factors were not important with regard to the choice of archetype within groupings.

3.26 The typology creation, however, only partially took account of size of dwellings and energy efficiency. Therefore, archetypes were chosen by selecting the dwellings with the median annual energy consumption[29] within a typology grouping. This ensured that variations in both size of dwellings and energy rating (within EPC band) were accounted for in the choice of archetypes.

3.27 Note that some archetype groups contained (and were therefore defined by) a single dwelling in the SHCS data. Similarly, a number of archetype groups covered only two cases in the SHCS data. For these groupings, a case was chosen at random to represent the archetype.

3.28 To confirm that the 355 chosen archetypes reflected the target stock, the prevalence of nine characteristics among these were compared with the prevalence in the target stock overall. Table 3.3 shows the results.

3.29 Overall, there was a very good match. As would be expected, there was very little difference with regard to characteristics that formed part of the typology creation. There was also little difference with regard to most characteristics that had not been part of the typology creation. For example, 31% of all dwellings in the target stock had rooms in the roof. In comparison, 32% of the archetypes (weighted to the target stock) had rooms in the roof. The archetypes were also very similar to the target dwellings overall with regard to un-insulated lofts and the proportion of dwellings that lacked mains gas.

Table 3.3: Comparison of prevalence of selection characteristics among all dwellings in the target stock and the modelled archetypes

Characteristic Among
all target dwellings (weighted) (N=1,786)
Among
modelled archetypes (weighted to target population)
Archetypes with characteristic (unweighted)
Detached dwelling 45% 45% 157
Sandstone walls 29% 30% 100
Mains gas boiler 40% 40% 96
Pre-1919 46% 47% 153
Rooms in the roof 31% 32% 106
Loft in original dwelling but less than 250mm of insulation 60% 60% 198
No mains gas 51% 51% 219
More than 5 rooms 44% 41% 137
Single glazed 18% 13% 58

3.30 Single glazing was less common among the archetypes than in the target stock. This means that dwellings with single glazing were under-represented in the outputs of the improvement outputs by about 20,000 (5% of the target stock). Therefore, the modelled outputs may have under-estimated the number of properties where double, secondary or triple glazing would be part of the package of cheapest measures to get these properties up to the various scenario bandings.

3.31 Dwellings with more than 5 rooms were also under-represented. However, because the archetype dwelling was chosen based on the median energy consumption, mid-size dwellings were more likely to be over-represented among the archetypes while very small and very large size dwellings will be likely to be slightly under-represented. Overall, the net effect on the modelled outputs is likely to be minimal.


Contact

Email: Silvia Palombi