Following Eq.(3), Figure 3 shows the decomposition of changes in emissions induced by buildings energy services across the four strategies identified above. Figure 3a clearly shows that the large growth in useful energy of 160% in the ‘Baseline’ scenario between 2015 and 2050, which reflects the improvement of living standards, exerts an important pressure on emissions (+127%) which is largely compensated by the progress in efficiency (-85%). Overall however, emissions increase by 54%. When only efficiency policies are put in place (Figure 3a, ‘EG’), improved efficiency almost entirely compensates for the impact of useful energy demand growth that would otherwise double emissions. The influence of supply-side factors is much lower than that of demand side factors on the change in emissions in the ‘Baseline’ and ‘EG’ scenarios between 2015 and 2050.
The opposite is true when considering the change in 2050 emissions between the ‘Baseline’ and 1.5°C scenarios. Demand side factors account at most for a decline in emissions of 18%, while measures decreasing the carbon content of energy contribute as much as a 74% decline in buildings emissions in the ‘1.5°C-EG’ scenario. That is, in this scenario, reducing the carbon content of energy contributes a share of 81% of the total 91% reduction in emissions. According to the decomposition, the 31% reduction of final energy demand in ‘1.5°C-EG’ compared to ‘Baseline’ translates into an 18%
emission reduction. As the scenario ‘EG’, which only removes market failures in efficiency markets, 146 Chapter 5 Decarbonising buildings energy services
does not envisage any policies targeted at CO2 emissions, fuel switching and supply decarbonisation do not contribute at all to the decline in emissions in this scenario. Accordingly, the achieved emission reductions correspond to the level of energy demand reductions (-16% and -17%,
respectively). A strategy that exclusively consists in removing barriers to energy efficiency is thus very unlikely to achieve emission reductions sufficient for the Paris climate targets.
Figure 3 Decomposition of direct and indirect emissions from buildings energy demand between 2015 and 2050 (a) for the ‘Baseline’ and ‘EG’ scenarios (see Supplementary Note 16 for the decomposition of other scenarios) and between the ‘Baseline’ scenario and other scenarios (b). Negative emissions from carbon capture and storage have been excluded from the accounting. The percentage values refer to the grey bars.
Looking into the details of the individual factors of Eq(1) (Figure 4), we observe that while the reduction in useful energy demand is limited to 10% and the reduction in energy intensity to 23% — leading to an aggregated drop in final energy demand of 31%—, the drop in the emission intensity factor reaches almost 90% in the ‘1.5°C’ and ‘1.5°C-EG’ scenarios. Several energy carriers can almost fully decarbonise (Figure 5): electricity, district heating, and solids via biomass. Residual emissions therefore stem to a large extent from the remaining demand for gases, whose share is still 11% in the
‘1.5°C-EG’ scenario (Figure 6). By 2050, the share of electricity is already above 60% in the ‘Baseline’
and rises to 75% in the most ambitious scenario. Accordingly, 74% of all 2050 buildings emissions
5.4 Contributions of demand and supply 147
stem from electricity in the ‘Baseline’ (Supplementary Note 14), making the decarbonization of electricity a fundamental requirement to reducing buildings sector emissions.
Figure 4 Changes across scenarios in the individual factors explaining the decline in buildings emissions (Eq(1)).
Final to Useful energy intensity is below one mostly because of air conditioners and heat pumps whose intensity can decrease below one, and to some extent to the treatment of appliances efficiency improvements (Supplementary Note 8).
Figure 5 Emission intensity of energy carriers in buildings. The red line shows the emission intensity in 2015.
The emission intensity of solids reflects the fact that an important share of solids comes from biomass which is accounted with no emissions here. The increase of the emission intensity for liquids and solids in ‘Baseline’ and
‘EG’ are due to a higher share of coal in solids while biomass is reduced, and a gradual adoption of coal-to-liquids (Supplementary Note 17).
148 Chapter 5 Decarbonising buildings energy services
Figure 6 Shares of buildings energy carriers in 2015 and 2050.
Discussion
The main insight from our study is that it is essential to adopt a full energy system perspective when assessing the potential for emission reductions in buildings: not only energy demand reductions but also carbon intensity reductions are important elements of an emission reduction strategy for the buildings sector in line with the Paris Climate targets. Importantly and despite the fact that our scenario is among the most ambitious 1.5°C scenarios in terms of energy demand reductions, it is the reduction of the carbon content of energy that accounts for 81% of the emission reductions in the sector compared to a baseline scenario without policy intervention. The remainder is explained by energy demand reductions. Focusing exclusively on energy demand reductions therefore overlooks more than 80% of the decarbonisation of buildings energy services. In particular, because of the almost entire decarbonisation of certain energy carriers like electricity or district heating, the
strategy of switching fuels has a high leverage to decrease emissions. To illustrate, holding everything else equal, a decrease in energy demand by one further percent in the ‘1.5°C-EG’ scenario would
5.5 Discussion 149
decrease emissions by only 12 MtCO2/yr (24 MtCO2/yr if all the decrease happens in space heating), because more than half of the saved energy would already be emissions-free. By contrast, raising the share of electricity or district heating by one percentage point at the expense of gas would decrease emissions by 65 MtCO2/yr in ‘1.5°C-EG’, as it would reduce gas demand, which is responsible for 62%
of residual emissions (Supplementary Note 14), by 10%.
Energy demand reductions, however, remain an important element: our scenarios show that policies alleviating the efficiency market failures raise the opportunities to decrease energy demand cost-effectively. In addition, efficiency improvements greatly moderate emission increases in the baseline, and strongly reduce the demand for decarbonised energy carriers in the ‘1.5°C-EG’, thus limiting the externalities in form of land and resource use35. But energy demand reductions, beyond their direct impact on emissions, are also important in their interactions with the other strategies to decrease emissions, especially electrification and fuel switching. For instance, by improving the insulation of buildings envelopes, the temperature of radiators can be reduced and the efficiency of heat pumps increased, further enabling the penetration of heat pumps. Higher energy efficiency also impacts the economics of district heating or the peak demand of electricity. It should be noted, however, that our understanding of the interactions and synergies between energy efficiency and fuel switching remains limited, especially in large-scale models, and thus is an important domain for future research.
Though significant, the reductions in energy demand presented in this study fall short of those showed in the most ambitious low energy demand scenarios published recently26,36. The difference is striking when comparing the scenario results with the historical energy demand. Compared with 2015, the scenario with the highest demand ambition presented in this study displays only a 3%
decrease in demand by 2050. In Grubler et al (2018)36 by contrast, buildings energy demand falls by 46% compared to historical values. The gap between these results derives primarily from the different perspectives adopted to address the topic of energy demand. Here, we were concerned with the optimal economic response to both efficiency market failures and climate change. The very ambitious scenarios rely instead on deep shifts in technologies, social norms, cultures, and tastes26 for which the political tools remain to a large extent unexplored.
This paper follows recent efforts to improve the representation of demand-side policies in integrated assessment models to include policies that go beyond carbon pricing37. The improvements in the modelling and policy representation however come with some limitations. To depict the energy efficiency market failures and behavioural barriers, we applied the concept of implicit discount rates.
The latter is, however, only an imperfect approximation for a variety of drivers explaining the divergence between observed and seemingly optimal behaviours. In addition, we deliberately kept consumers’ service demand constant across scenarios (see Methods). Only the choice of technology options to fulfil this demand was left to the model. This however forgoes two important effects. The service demand might decrease in response to carbon pricing, because the latter increases the costs of energy services. On the other hand, the service demand might increase in response to energy efficiency policies, because the higher efficiency reduces the costs of energy services —the rebound effect. Our choice to fix the energy service demand probably overestimates the reaction to energy efficiency policies and underestimates the reduction of demand following carbon pricing.
Despite these caveats, our study clearly shows that assessing the decarbonisation of buildings energy services requires an investigation of all factors behind buildings emissions, i.e. of both demand and
150 Chapter 5 Decarbonising buildings energy services
supply. Essential and dominant dynamics behind buildings decarbonisation are overlooked when focussing mainly on energy demand reductions. Integrating strategies like fuel switching into the analysis of buildings decarbonisation is therefore fundamental for future research, as they could reduce residual emissions from this sector and therefore limit the need for debated negative emissions technologies.
Methods
Description of the REMIND model
In this study, we use the global energy-economy model REMIND38,39. REMIND includes
representations of the economic, energy and climate systems. It is used to conceive of possible pathways to curb climate change, but also to assess the social and environmental implications of these pathways.
In REMIND, the macroeconomic output is a function of the inputs labour, capital and aggregated energy services. The aggregated energy services derive from the energy consumption in three sectors: buildings, industry and transport. Each sector requires final energy carriers to provide the sector-specific energy services. The economic output is used for consumption, trade, investments into the macroeconomic capital stock, and energy system expenditures.
The energy supply system provides the energetic inputs required by the economic system. The energy supply explicitly represents vintage capital stocks for more than 50 conventional and low-carbon energy conversion technologies and tracks energy flows from primary through secondary to final energy.
REMIND determines the optimal allocation of resources to maximise the welfare of the representative agents from 12 global regions until 2100. As a result, REMIND projects energy demand trajectories and their disaggregation across sectors and energy carriers. It also projects pathways for primary and secondary energy production, accounting for the emissions ensued by energy supply and consumption.
The macroeconomic and the energy system modules are hard-linked via final energy demand and costs incurred by the energy system. Energy production and final energy demand are determined by market equilibrium —that is, between marginal utility and marginal costs of energy use.
Buildings module general structure
The buildings module in REMIND depicts the energy demand at the final and useful energy levels for four service categories: appliances-lighting, cooking-water heating, space heating and space cooling.
Cooking and water heating are bundled because both consume similar energy carriers and the options to decrease energy consumption are assumed to be similar in comparison with the other services. Water heating is quite similar to space heating as for the options available to increase energy efficiency, but it is separated from space heating in the model because space heating is strongly influenced by the quality of buildings’ envelopes, while this is not the case for water heating.
Importantly, we introduce in the model the opportunity to substitute between energy consumption and energy end-use capital (on the definition of end-use capital, please refer to Supplementary Note
5.6 Methods 151
7). The production of an energy service is depicted as a function of end-use capital and energy use.
For a given level of energy service, the model can increase the level of end-use capital in order to decrease the amount of energy consumed. Thereby the model can define endogenously the level of efficiency in response to the evolution of capital and energy prices. There are three such substitution opportunities in the buildings sector: for the buildings’ envelope, for air conditioners, and for
appliances and lighting.
Formally, the trade-off is represented through a Constant Elasticity of Substitution function (CES).
The degree of substitutability is determined by the elasticity parameter of the CES function. The elasticity of substitution, a crucial parameter for the energy efficiency potential, is calibrated on technological data (see Supplementary Note 4). In the long run, technological developments also react to price changes, leading to a greater reactiveness of the system to price variations. Because endogenous technological development for end-use technologies is not represented in the model, we assume that the long run elasticity of substitution increases over time (Supplementary Note 4).
For space heating and water heating – cooking, the model can decide upon the shares of conversion technologies employed. Each conversion technology distinguishes itself through its capital costs, efficiency and the energy carrier used. However, the model cannot endogenously raise the efficiency of individual technologies.
The energy efficiency gap and the implicit discount rate
In the late seventies, Hausman 4 observed that in order to explain purchasing habits of energy consuming appliances with the net present value of these appliances, it was necessary to apply a discount rate which was much higher than observable interest rates. Implicit discount rates are the discount rates that make observed purchasing decisions coherent with decisions taken according to the net present value of alternatives. As energy efficient technologies have lower operating costs but higher initial capital costs, high discount rates give inefficient technologies a competitive edge and
“explain” why people tend to prefer these technologies.
As described in Schleich 28, the implicit discount rate does not only represent preferences and short-termism or risk aversion of individuals, it also stands for a host of barriers which comprises market failures and behavioural aspects, and which are not easy to represent explicitly. Though they are an imperfect representation of barriers 28,40, implicit discount rates have already been used in several models to represent households and organisations’ behaviour regarding energy efficiency
investments 41–43.
It should be noted that the implicit discount rate, and its use in energy models, should not be confused with the private discount rate, and that the discussion on the use of implicit discount rates should not be confused with the debate around social and private discount rates 44,45. Indeed, except in the cases where market and behavioural barriers are explicitly modelled in energy models, the discount rate constitutes an indirect mean of representing these barriers.
In REMIND, the macroeconomic discount rate is endogenously computed by the model. To model the implicit discount rate in buildings, we impose a tax on the end-use capital and recycle the tax
revenues in a lump sum fashion. This pro rata tax increases the macroeconomic discount rate additively to mimic an end-use specific implicit discount rate. For instance, if the macroeconomic discount rate is 7% and the tax on end-use capital is 10%, the full discount rate on end-use capital
152 Chapter 5 Decarbonising buildings energy services
will be 17%. The target implicit discount rates (Table 3) are consistent with the lower bound of the literature to reflect the effect from energy efficiency policies that have already been implemented 46. Variable Increase in the discount rate
(tax ratio on the efficiency capital)
Target implicit discount rate (based on an assumption of a 7%
endogenous discount rate)
Insulation 5 pp 12%
Space heating 5 pp 12%
Space cooling 5 pp 12%
Water heating and cooking 5 pp 12%
Appliances and Lighting 20 pp 27%
Table 3: Values for the implicit discount rates. The abbreviation pp stands for percentage points. The endogenous discount rate is approximately 7% in equilibrium. Considering this rate, we taxed the efficiency capital by a tax rate which will yield an overall endogenous discount rate for these capital stocks close to implicit discount rates from the literature.
Buildings long-standing infrastructure
Buildings constitute some of the longest-lived infrastructure. The average life expectancy of a building is counted in decades and only a low percentage of buildings is renovated every year. Even dramatic energy price changes or strong policies targeting insulation efficiency are unlikely to result in the rapid refurbishment of the whole buildings stock. As a result, the efficiency of buildings’
envelope can only evolve slowly. Investment decisions in the short term have long term impact on the energy consumption of buildings’ users.
We include this inertia into the model through a putty-clay formulation. At each period, a decision is taken as for the ratio between energy consumption and efficiency investments for new and
renovated buildings. This choice on the marginal stock is represented through a CES function (4). For the sake of numerical simplicity, the new investments are added to the capital of the next period, without differentiating between the vintages within the standing capital stock (8). The energy services produced by a combination of energy demand and capital investments are then depreciated at the same rate as the corresponding buildings (6)-(8).
Δ𝑆𝑆𝑒𝑒𝑆𝑆𝑣𝑣𝑆𝑆𝑐𝑐𝑒𝑒𝑡𝑡= [(𝛼𝛼𝐹𝐹Δ𝑈𝑈𝑡𝑡)𝜌𝜌+ (𝛼𝛼𝐾𝐾Δ𝐾𝐾𝑡𝑡)𝜌𝜌]𝜌𝜌1 (4)
𝜌𝜌= 1−1
𝜎𝜎 (5)
𝑆𝑆𝑒𝑒𝑆𝑆𝑣𝑣𝑆𝑆𝑐𝑐𝑒𝑒𝑡𝑡= (1− 𝛿𝛿)𝑆𝑆𝑒𝑒𝑆𝑆𝑣𝑣𝑆𝑆𝑐𝑐𝑒𝑒𝑡𝑡−1+Δ𝑆𝑆𝑒𝑒𝑆𝑆𝑣𝑣𝑆𝑆𝑐𝑐𝑒𝑒𝑡𝑡
(6)
𝑈𝑈𝑡𝑡= (1− 𝛿𝛿)𝑈𝑈𝑡𝑡−1+Δ𝑈𝑈𝑡𝑡 (7)
𝐾𝐾𝑡𝑡 = (1− 𝛿𝛿)𝐾𝐾𝑡𝑡−1+Δ𝐾𝐾𝑡𝑡 (8)
Where αi is the factor-augmenting efficiency parameter of the variable i; ρ is a parameter related to the elasticity of substitution σ (5); t the time index; Service the energy service; E the energy
consumption; K the end-use specific capital stock.
5.6 Methods 153
Decomposition of effects across factors
Emissions from buildings energy services result from the level of useful energy demand, the conversion efficiency between final and useful energy, the shares of energy carriers as well as the carbon content of each energy carrier. To decompose the changes in emissions across these interacting factors, we proceed in two steps. First, we attribute the effects to the multiplicative factors of Eq(1) according to the methodology of ref 47, and then we decompose the effect of the emission intensity of fuels across fuel switching and supply decarbonisation.
In the first step, we rewrite the change in emissions as follows:
𝐶𝐶𝑂𝑂2=𝑈𝑈𝑈𝑈×𝐹𝐹𝑈𝑈 𝑈𝑈𝑈𝑈×𝐶𝐶𝑂𝑂2
𝐹𝐹𝑈𝑈 =𝑈𝑈×𝐼𝐼×𝐶𝐶 (9)
Δ𝐶𝐶𝑂𝑂2=𝐶𝐶𝑂𝑂2𝐹𝐹− 𝐶𝐶𝑂𝑂2𝐼𝐼 = (𝑈𝑈𝐼𝐼+Δ𝑈𝑈)(𝐼𝐼𝐼𝐼+Δ𝐼𝐼)(𝐶𝐶𝐼𝐼+Δ𝐶𝐶)− 𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐶𝐶𝐼𝐼 (10) With F for final state and I for initial state. The initial state is 2015 in Figure 3a and it is the year 2050 of the ‘Baseline’ scenario in Figure 3b. Eq. (9) can be reorganised as
Δ𝐶𝐶𝑂𝑂2= 𝑢𝑢𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒+𝑐𝑐𝑐𝑐𝑐𝑐𝑣𝑣𝑒𝑒𝑒𝑒𝑒𝑒+𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒 (11) With
𝑢𝑢𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒=Δ𝑈𝑈× (𝐼𝐼𝐼𝐼𝐶𝐶𝐼𝐼+1
2𝐼𝐼𝐼𝐼Δ𝐶𝐶+1
2Δ𝐼𝐼𝐶𝐶𝐼𝐼+1
3Δ𝐼𝐼Δ𝐶𝐶) (12)
And similarly for 𝑐𝑐𝑐𝑐𝑐𝑐𝑣𝑣𝑒𝑒𝑒𝑒𝑒𝑒 with Δ𝐼𝐼, and 𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒 with Δ𝐶𝐶.
We then need to decompose Δ𝐶𝐶, the effect of carbon intensity, between fuel switching (fs) and the supply decarbonisation (sd). Following Eq.(2), we write:
Δ𝐶𝐶=𝐶𝐶𝐹𝐹− 𝐶𝐶𝐼𝐼= � 𝐶𝐶𝑒𝑒𝑒𝑒𝐹𝐹𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝐹𝐹 𝑒𝑒𝑒𝑒
− � 𝐶𝐶𝑒𝑒𝑒𝑒𝐼𝐼 𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝐼𝐼 𝑒𝑒𝑒𝑒
=� 𝐶𝐶𝑒𝑒𝑒𝑒𝐹𝐹(𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝐹𝐹 − 𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝐼𝐼 𝑒𝑒𝑒𝑒
) +�(𝐶𝐶𝑒𝑒𝑒𝑒𝐹𝐹 − 𝐶𝐶𝑒𝑒𝑒𝑒𝐼𝐼 )𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝐼𝐼
=Δ𝐶𝐶𝑒𝑒𝑓𝑓+Δ𝐶𝐶𝑓𝑓𝑠𝑠 𝑒𝑒𝑒𝑒
(13)
Finally, we use Δ𝐶𝐶𝑒𝑒𝑓𝑓 and Δ𝐶𝐶𝑒𝑒𝑠𝑠, to decompose 𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒 into 𝑓𝑓𝑠𝑠𝑒𝑒𝑒𝑒𝑒𝑒 and 𝑠𝑠𝑑𝑑𝑒𝑒𝑒𝑒𝑒𝑒:
𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒=Δ𝐶𝐶×𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒
Δ𝐶𝐶 = Δ𝐶𝐶𝑒𝑒𝑓𝑓×𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒
Δ𝐶𝐶 +Δ𝐶𝐶𝑓𝑓𝑠𝑠×𝑐𝑐𝑐𝑐𝑆𝑆𝑏𝑏𝑒𝑒𝑒𝑒𝑒𝑒
= 𝑓𝑓𝑠𝑠𝑒𝑒𝑒𝑒𝑒𝑒+ 𝑠𝑠𝑑𝑑𝑒𝑒𝑒𝑒𝑒𝑒 Δ𝐶𝐶
Baseline scenario
Final or useful energy projections for buildings energy service can best be produced with a model covering the demand for living space, quality of buildings’ envelope, the heating and cooling degree days, etc. Because of its numerical complexity, REMIND does not include all these variables. Instead, we use the model EDGE to design buildings energy demand projections 2,26, and then calibrate the REMIND baseline scenario to meet these projections (Supplementary Note 3). In the Baseline scenario, no policy action is taken and no steps are taken to reduce the energy efficiency gap, i.e. the implicit discount rates stay constant over time and the carbon price is set to zero.
154 Chapter 5 Decarbonising buildings energy services
Efficiency gap scenario
The ‘Efficiency gap’ scenario mimics the implementation of non-pricing policies aiming at reducing the barriers to energy efficiency, and therefore reducing the implicit discount rates. Such policies
The ‘Efficiency gap’ scenario mimics the implementation of non-pricing policies aiming at reducing the barriers to energy efficiency, and therefore reducing the implicit discount rates. Such policies