Evolving Green Premiums: The Impact of Energy Efficiency on London Housing Prices over Time

Abstract

As climate policy and energy costs increasingly influence housing markets, understanding how energy efficiency is capitalized into home prices has become a critical question for both researchers and policymakers. While prior studies confirm the existence of a green premium—the price advantage of more energy-efficient homes—less is known about how this premium evolves over time in response to shifting regulations, awareness, and market conditions. This study provides new empirical evidence on the dynamic valuation of energy efficiency in the London housing market between 2013 and 2021. Using a repeat-sales framework, we isolate within-property price changes and examine how energy performance is capitalized over time. We find that the green premium associated with higher current energy efficiency strengthened steadily, rising from statistically insignificant levels in 2013 to approximately 0.47% per Energy Performance Certificate (EPC) point by 2021. Meanwhile, the price penalty for a large efficiency gap, reflecting unrealized upgrade potential, narrowed substantially in 2020 and 2021, indicating a marked reduction in buyers’ aversion to less efficient homes. This study adds a new dimension to the green premium literature. It provides empirical evidence that the relationship between energy efficiency and housing value is not static, but responsive to regulatory, economic, and social changes. By tracking year-by-year changes in London, our analysis offers insight into how quickly market preferences adjust and how interventions like minimum efficiency standards translate into property values. This enriched understanding moves the field beyond the question of whether a green premium exists, to how and why it evolves.

1. Introduction

Energy efficiency in housing has emerged as a critical concern at the intersection of climate policy and real estate markets. Residential buildings are a major source of greenhouse gas emissions, so improving home energy performance is essential to achieve sustainability goals. In parallel, energy-efficient homes offer reduced utility costs and improved comfort, attributes that can translate into higher market valuations as buyers increasingly factor in sustainability and cost savings.

A growing body of research indicates that more energy-efficient homes do indeed command a green premium (also referred to as an energy efficiency premium)—that is, a higher sale price compared to otherwise similar but less efficient homes. Empirical studies across Europe (including the UK) consistently find positive price effects for higher energy efficiency. Many studies use Energy Performance Certificate (EPC) ratings as the measurement of energy efficiency. EPC was introduced in 2007–2008 as part of EU climate directives and provides a standardized measure of a dwelling’s energy efficiency. An EPC assigns each home an efficiency rating on a scale from A (most efficient) to G (least efficient) and is legally required when a property is constructed, sold, or rented. A recent review of 68 studies concludes that each one-letter improvement in a home’s EPC band is associated with roughly a 1–3% increase in its sale price. This implies that buyers are willing to pay a premium for efficiency, likely reflecting expectations of lower energy bills and improved comfort. However, the existing literature has largely treated this green premium as a static phenomenon. Little is known about how the premium for energy efficiency may evolve over time. A key gap in our understanding is the following question: are buyers placing greater importance on energy efficiency now than they did a few years ago?

This paper addresses this gap by examining the temporal dynamics of the green premium in the London housing market over 2013–2021. Specifically, we investigate two main research questions:

a.

How has the price premium associated with energy efficiency in London’s housing market evolved from 2013 to 2021?

b.

Has the market’s valuation of unrealized energy-saving potential changed over the same period?

We hypothesize that both the green premium and the value attached to retrofit potential have increased over time. This expectation is motivated by several converging trends: homebuyers’ preferences are shifting in favor of sustainable, lower-cost homes; policymakers have sent stronger signals that raise awareness of EPC ratings; and energy prices have generally risen, making energy savings more financially salient. Together, these factors suggest that efficient properties should enjoy a growing pricing advantage, and that even properties with lower current efficiency but high improvement potential may be viewed more favorably in later years than in earlier years. To answer our research questions, we employ a repeat-sales econometric design with year-specific interactions for energy efficiency variables. In essence, we track individual properties across multiple transactions and estimate how the price impact of energy performance (EPC efficiency score and efficiency gap) varies by year. This approach controls for unobservable property traits by differencing within the same home, and it allows us to isolate the year-by-year evolution of the green premium. Using a panel of London home sales from 2013 to 2021, a period of active market turnover and intensifying climate policy focus, we find clear evidence of changing valuations. Our study provides novel quantitative evidence on how the market capitalization of residential energy efficiency evolves over time.

2. Literature Review

2.1. Empirical Evidence of Green Premiums in Housing Markets

A growing body of empirical research has examined whether green premiums—higher prices for more energy-efficient homes—exist. Many studies have found a statistically significant, albeit modest, premium associated with superior energy performance labels. Early evidence from European markets confirmed that buyers and renters place a positive, significant value on better energy ratings, all else being equal [1,2,3,4,5,6,7,8]. For example, an influential study in the Netherlands found that compared to a median “D”-rated dwelling, houses with an “A” label sold for roughly 10% more on average, with progressively smaller premiums for “B” (~5.5%) and “C” (~2.5%) ratings and corresponding discounts for the lowest ratings [1]. This study provided some of the first clear evidence that energy efficiency is capitalized into house prices. In England, Fuerst et al. report that homes rated A or B sold for about 5% more, and C-rated homes ~1–2% more, relative to baseline D-rated homes [9]. A green premium has also been observed later in the rental market. Even after controlling for property attributes, higher EPC-rated rentals achieved slightly higher and statistically significant rents, while the lowest-rated properties (e.g., G) faced price penalties [10]. Despite variations in energy rating systems and administrative frameworks across jurisdictions, consistent evidence of a green premium has been observed in both rental and sales markets in regions such as North America, Latin America and Asia [11,12,13,14,15,16]. Beyond transaction price analyses, several studies investigate the green premium from the buyers’ perspective by directly examining willingness-to-pay with surveys and similarly identify its presence. [17]. Overall, the literature documents a consistent directional effect: more energy-efficient homes generally tend to transact at higher prices or rents. But the magnitude of the green premium varies across studies and contexts [18]. Specifically, the price premium associated with higher energy efficiency ratings, such as A or B, varies considerably across studies, ranging from about 10% to as high as 30% [19]. Such variation can be partly explained from these two aspects:

• Contextual Differences: Markets with higher baseline fuel costs or greater climate-driven energy needs tend to show stronger willingness-to-pay for efficiency. A study shows that Turin’s buyers paid close attention to the EPC label, reflecting the greater heating needs (higher heating degree days) and fuel use there, whereas Barcelona’s buyers placed more value on features like air conditioning and pools to cope with hot summers [20]. Additionally, evidence from the UK indicates higher relative premiums in regions where the house price-to-earnings ratio is higher or housing tenure is longer [21]. Property type also matters: flats and terraced homes, which are often associated with more budget-conscious buyers or tenants, show higher percentage premiums than detached houses in some studies [22]. However, in contexts where energy is cheap or incomes are high, the premium can shrink. A study in Oslo (Norway) found virtually no price effect from EPC ratings, concluding that homebuyers placed minimal weight on energy efficiency compared to location, size, and other attributes [23]. Similarly, research in Sweden’s housing market found no significant price premium attributable to the EPC label itself, aside from the value of specific efficiency-related attributes (like better insulation or windows) [24]. This null result contrasts with the notable premiums observed in other countries, and it underscores how abundant cheap energy, cost of retrofitting and other financial factors can neutralize the willingness to pay for efficiency [25,26].
• Influence of Information and Policy: Several studies highlight the role of disclosure and standards in unlocking the green premium. Ghosh et al., exploiting an “information shock” policy that made EPC ratings more salient at the point of sale, provide causal evidence that homebuyers will pay more for better-rated homes. They estimate a 1–3% price premium per EPC letter upgrade at the national level, and a somewhat higher 3–6% premium in the London market specifically [21]. Government policies can also create direct financial incentives: for example, from 2018 the UK’s Minimum Energy Efficiency Standards (MEES) regulation made it unlawful to rent out properties below EPC “E”, effectively penalizing the worst-rated homes and likely increasing investors’ attention to energy performance [27]. Such regulatory shifts are expected to be capitalized into prices. Inefficient properties facing mandatory retrofits or rent restrictions should sell at discounts, while efficient ones may enjoy enhanced demand. Correspondingly, when the policy was patchily enforced, early research found weaker or inconsistent effects. Early evidence on behavioral responses to MEES reveals that the early compliance was “too easy” and landlords could obtain exemptions or make minimal changes, so the policy’s immediate impact on the market was muted [28]. However, the study also found MEES beginning to change market practice, for instance, some landlords started adding clauses or seeking greater control over tenant behavior in leases to ensure compliance. These insights suggest that policy alone may not immediately transform market values without strong enforcement, but it has started to shift investor awareness and behavior. Indeed, industry observers have posited that the value attached to energy efficiency will rise over time as climate policies tighten, and that market premiums could grow if government incentives for green homes strengthen.

In summary, the empirical literature supports the existence of a green premium in housing markets, especially in the UK. Energy-efficient dwellings generally attract higher prices or rents than otherwise comparable inefficient dwellings, though the premium is typically modest and contingent on market context.

2.2. Gaps in Research: Temporal Dynamics and Policy Context

While the cross-sectional evidence for green premiums is well documented, a notable gap in the literature mentioned by some studies is the limited understanding of temporal dynamics, which is how the pricing of energy efficiency in housing might change over time. Most studies implicitly assume a constant effect of energy ratings on price within their study period, or they pool data across years without investigating changes in the premium. Only a handful of papers have attempted to explicitly test whether the green premium is growing, shrinking, or fluctuating as market conditions evolve. Marmolejo-Duarte and Chen [29], using listing price data from 2014 and 2016, apply pooled spatial regression analyses and find that the market premium for energy efficiency increased over time in Spain’s second-largest urban agglomeration. Another study analyzes external appraisals in England and the Netherlands across two years, showing that while energy efficiency was not reflected in assessed values in 2012, by 2015 appraisers had begun to incorporate significant green premiums in the Netherlands and discounts for lower-rated dwellings in England [30].

In addition, several review studies synthesize individual findings to examine whether a consistent trend in the green premium emerges over time. Cespedes-Lopez et al. [31] conducted a systematic quantitative review (meta-analysis) of dozens of studies on EPC-related price effects up to 2019. The meta-analysis confirms that energy-efficient homes do sell for a premium, but compared with the premium from the 2014 meta-study, their results show somewhat lower effects. This could suggest that the green premium had stabilized or even attenuated slightly in later years, or simply that more diverse markets, some with low premiums, were studied. The meta-analysis highlights a lack of longitudinal studies in the literature, which is later reinforced by a recent scoping review of European studies. The European review study confirms that findings on any change in the premium over time are mixed and that this question remains under-explored [32]. The authors note that existing longitudinal studies are few in number, and they call for “future research should make further efforts to explore impact of EPCs on price change” with more recent data and refined methods [32]. In short, it is not yet well resolved whether markets are attaching greater value to energy efficiency now than, say, five or ten years ago, or how quickly any such adjustment is occurring.

This gap in knowledge is especially important in light of the rapidly shifting policy and market context of the past decade. A study explicitly examines how the EPC price effect changed before and after the introduction of EPC in Barcelona and demonstrates that the green premium grew significantly in just a few years [29]. The study supports the idea that the trajectory of green premiums is upward when catalyzed by policy enforcement and public consciousness. It also exemplifies the importance of longitudinal analysis. In the UK, the period since 2015 has seen a series of developments that could plausibly affect the green premium. Regulatory changes have raised the floor for acceptable energy performance, notably the 2018 MEES regulation in England and Wales, which effectively rendered the worst-performing rentals unlettable. As discussed above, this kind of policy can create an immediate price effect and may also increase general awareness of energy efficiency in the property market. Although there is a lack of temporal studies, we find that two separate studies about green premium in Belfast at two different years indicate a slightly stronger recognition of EPC ratings from 2015 to 2020 [33,34]. Besides EPC-related policies, the UK government also has set ambitious climate targets, like the net-zero by 2050 commitment, and launched public campaigns and subsidy programs for home retrofits which might also influence consumer attitudes. Meanwhile, energy prices have fluctuated substantially, especially with the global energy crisis unfolding in 2021–2022. Higher fuel prices make the cost savings from an efficient home more economically significant, potentially boosting buyers’ willingness-to-pay. Indeed, analysts have hypothesized that the surge in energy costs could translate into greater demand for efficient homes, although empirical evidence is just beginning to emerge [35]. At the same time, the baseline energy efficiency of the housing stock has been improving, which could either increase the premium by making efficiency more normative and expected, or decrease it if efficient homes become commonplace. The net effect of these opposing forces is unclear.

In summary, there is a strategic research need to move beyond static estimates of the green premium and examine how it unfolds over time in response to external triggers like policy implementation, energy price shocks, or shifts in public sentiment. The literature to date offers limited insight into the following question: Has the pricing premium for energy-efficient homes strengthened as climate policies and market awareness have intensified? Early evidence provides hints but no consensus. Future market dynamics could differ markedly from past averages, so a temporal approach is required to capture any momentum or structural change in the valuation of energy efficiency.

3. Methodology and Data

3.1. Theoretical Framework: Hedonic Pricing and Energy Efficiency

Housing is viewed through the lens of hedonic pricing theory as a differentiated good comprising a bundle of attributes that homebuyers value [36,37]. In this framework, the sale price of a property is determined by its characteristics, such as location, size, amenities, energy efficiency, etc., each contributing an implicit price reflecting buyers’ marginal willingness-to-pay. We formalize housing $\mathrm{H}$ as a function of observed quality (${\mathrm{Q}}_o$), unobserved quality ($Q_u$), and energy efficiency ($\mathrm{EE}$). Here $Q_o$ includes measurable characteristics, such as floor area, room count, location, and age, while $Q_u$ captures latent features, such as the build quality, architectural appeal, and so forth, that are not directly observed by our data. Energy performance EE, which is usually evaluated based on energy consumption, is one component of $\mathrm{H}$ that contributes to utility and price. Under this theoretical construct, the present value of expected energy cost savings (plus any comfort or environmental benefits) from better energy efficiency should be capitalized into the asset price of the home. This logic is consistent with the user-cost theory of housing: a house with lower running costs, like heating, cooling, etc., effectively offers a lower total cost of ownership, enabling buyers to pay more upfront for the asset. Energy efficiency enters the buyer’s utility function by signaling these lower future energy expenditures and possibly higher comfort, thereby increasing the buyer’s willingness-to-pay. Therefore, the market price reflects the bundle of benefits the home offers, including energy cost savings and environmental amenities delivered by better energy efficiency.

While a standard cross-sectional hedonic regression can estimate the implicit price of energy efficiency by controlling for observable characteristics, it may suffer from omitted variable bias in practice. If energy-efficient homes systematically differ in unobserved qualities from inefficient homes, a simple hedonic model could misattribute the price differences to energy performance when part of the premium is due to those unobserved factors. Researchers have indeed cautioned that estimated green premiums might be biased upward due to such omitted variables, including building quality or curb appeal that correlate with higher energy efficiency. For example, some studies argue that some of the price uplift attributed to efficiency may reflect broader renovations or locational advantages often found in efficient homes [5,29,38,39].

To address this concern, our study adopts a repeat-sales estimation strategy, leveraging the panel nature of housing transactions. The repeat-sales approach tracks the same property over time, thereby controlling for all time-invariant unobserved attributes by design [40]. By examining price changes for a given house between two sale dates, we difference out characteristics that do not change (the fixed $Q_u$ and constant aspects of $Q_o$ for that property), ensuring that “like is compared with like” in terms of intrinsic quality. This method greatly mitigates the omitted variable problem because any intrinsic quality or locational advantages of a particular house are present in both the initial and subsequent sale and thus cancel out when we look at the price difference. In existing studies, Fuerst et al. apply the repeat-sales framework in the UK context to verify that the EPC effect on appreciation remains positive even after controlling for house fixed effects [8]. In addition, by using a repeat-sales design, we also naturally correct for the quality-mix problem that plagues simple price trend analyses: because only properties with at least two transactions are analyzed, and each property is matched with itself over time, we avoid distortions arising from different types of homes selling in different periods. It is worth noting that the repeat-sales method assumes that a property’s latent quality remains approximately constant between sales. However, in reality, this may not strictly hold: an upgrade in energy efficiency is often part of a broader home improvement. For example, an owner might undertake a major renovation at the same time as improving insulation or heating systems. Such changes mean that $Q_u$ or even parts of $Q_o$ are not truly constant, which could confound the energy effect if ignored. We respond to this by augmenting the repeat-sales approach with controls for observable renovations, described below, and by carefully constructing the dataset to remove repeat sales with large upgrades. A further drawback of repeat-sales indices is sample-selection bias: they use only homes that trade at least twice during the sample period. If those ‘frequent sellers’ differ from the broader stock, the analysis result can misstate market-wide price [41]. Despite these considerations, the repeat-sales model is a powerful approach to net out time-invariant house heterogeneity and has been widely used in housing economics for constructing constant-quality price indices [40,42].

3.2. First-Difference Repeat-Sales Model Specification

We implement a first-difference repeat-sales regression to quantify the green premium in Greater London. The dependent variable is the change in the log sale price of property i between two transactions, $\Delta \ln P_{it}=\ln P_{i,t}-\ln P_{i,t_0}$. A log-price specification is chosen so that coefficients can be interpreted as approximate percentage changes in price, and to mitigate heteroskedasticity. On the right-hand side, our key explanatory variable is the change in energy efficiency between sales. We denote by $\Delta \mathrm{EE}$ the improvement (or deterioration) in the home’s energy efficiency score from EPC. $\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}$ thus captures movements of the property $i$ at time $t$. In addition to current efficiency, we introduce a novel variable to account for retrofit potential: the EPC report provides not only a current energy efficiency but also a potential efficiency if recommended cost-effective improvements are carried out. We define an “efficiency gap” as the difference between the potential and current rating, and include the change in this gap between sales, $\Delta \mathrm{E}{\mathrm{E}}^{\mathrm{gap}}$. This term measures how the unrealized energy efficiency potential evolves. For instance, if a home’s current energy efficiency score was 72 with a potential score of 81 at first sale, and by second sale the current score improved to 75 with the same potential score, the gap might shrink, and our model would capture this reduction. Intuitively, $\Delta \mathrm{E}{\mathrm{E}}^{\mathrm{gap}}$ lets us isolate the effect of closing the efficiency gap as distinct from simply having a higher score. A significant coefficient on this variable would indicate that buyers value not only the achieved efficiency level but also the remaining improvement potential (or lack thereof) in a home. Including both $\Delta \mathrm{EE}$ and $\Delta \mathrm{E}{\mathrm{E}}^{\mathrm{gap}}$ allows us to disentangle the price gain from actual efficiency improvements versus any price effects of a property’s latent efficiency opportunities.

In terms of controls, we account for several factors that could confound the estimated effect of energy efficiency. We first control for changes in key structural attributes. These include changes in floor area and the number of rooms between sales. In principle, for most properties these attributes remain constant; however, if an owner has extended the property or reconfigured the layout, such changes will be reflected in $\Delta X$. In addition, we include month-of-sale fixed effects $M$ to account for seasonal fluctuations in housing prices. To allow for heterogeneity in how energy efficiency is valued across different housing types, we also interact property type indicators $Z$ with both the energy efficiency change $\Delta \mathrm{EE}$ and the energy efficiency gap $\Delta \mathrm{E}{\mathrm{E}}^{\mathrm{gap}}$. These interaction terms permit the pricing effect of energy efficiency to vary by structural typology [10,43]. By including these controls, we ensure that the estimated coefficients on energy efficiency gains are not conflated with seasonal variation, structural differences across dwelling types, or major physical renovations.

A key extension of our model is the inclusion of interaction terms with year indicators to capture temporal variation in the green premium and the valuation of retrofit potential. Rather than assuming the effect of an efficiency upgrade is constant over the entire sample period, we allow it to vary by sale year. Specifically, we interact $\Delta \mathrm{EE}$ with a full set of year dummies. This yields year-specific coefficients ${\mathsf{\lambda }}_y$, so that an energy efficiency improvement in, say, 2015 might have a different impact on price appreciation than an identical improvement in 2020. Similarly, we interact $\Delta \mathrm{E}{\mathrm{E}}^{\mathrm{gap}}$ with year dummies to allow the market’s valuation of unrealized efficiency potential to evolve over time. Formally, if $D_y$ is an indicator for the second sale occurring in year y, we include terms like $\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{i}}\times D_y\right)$ and $\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{i}}^{\mathrm{gap}}\times D_y\right)$ for each year. These interaction terms flexibly capture any temporal trends in how buyers value energy performance improvements. The equation of the first-difference regression model is

$$\begin{array}{cc}\Delta \ln P_{it}& =\mathsf{\alpha }+{\displaystyle \sum _{y=2014}^{2021}}{\mathsf{\lambda }}_y{D}_y+{\displaystyle \sum _{y=2014}^{2021}}{\mathsf{\beta }}_y\hspace{0.17em}\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}\times D_y\right)+{\displaystyle \sum _{y=2014}^{2021}}{\mathsf{\gamma }}_y\hspace{0.17em}\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}^{\mathrm{gap}}\times D_y\right)\hfill \\ & +\delta \Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}+\theta \Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}^{\mathrm{gap}}+{\displaystyle \sum _m}{\mathsf{\mu }}_m{M}_m+{\displaystyle \sum _k}\left[{\mathsf{\eta }}_k\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}\times Z_k\right)+{\mathsf{\xi }}_k\left(\Delta \mathrm{E}{\mathrm{E}}_{\mathrm{it}}^{\mathrm{gap}}\times Z_k\right)\right]+\Delta X_{it}^{\prime }\mathsf{\varphi }+{\mathsf{\epsilon }}_{it}\hfill \end{array}$$

We estimate the model using Ordinary Least Squares (OLS). To ensure robustness, all input variables are converted to numeric types, and the model is run on a filtered subset where valid energy performance data is present for both sales. In terms of interpreting the model result, the dependent variable, $\Delta \ln P_{it}$, represents the logarithmic difference in sales prices of the same property i between two transactions. It is calculated as $\ln \left(P_{i{t}_2}\right)-\ln \left(P_{i{t}_1}\right)=\ln \left({\displaystyle \frac{P_{i{t}_2}}{P_{i{t}_1}}}\right)$, where $t_1$ and $t_2$ denote the earlier and later sale dates, respectively. This transformation allows us to interpret the outcome approximately as the percentage change in price between the two sales: $\Delta \ln P_{it}\approx {\displaystyle \frac{P_{i{t}_2}-P_{i{t}_1}}{P_{i{t}_1}}}$ . This approximation is accurate when the price change is small or moderate and provides a consistent basis for measuring within-property appreciation, net of time-invariant property characteristics.

3.3. Data Sources and Variable Definitions

This study combines two comprehensive data sources to analyze the “green premium”. The core housing transactions data come from HM Land Registry’s Price Paid Data (PPD), which records address-level residential property sales in England and Wales with near-universal coverage of sales. The second data source is the EPC dataset released by the UK Government. EPC became mandatory for properties when constructed, sold, or rented in UK in 2008, and it provides a detailed energy performance evaluation along with recommendations and property features. Merging these two datasets allows us to observe both the transaction price and the energy efficiency metrics for the same property. In our merged dataset, we include variables such as the number of habitable rooms, total floor area, current energy efficiency score, potential energy efficiency score, current EPC rating, and potential EPC rating from the EPC, alongside the sale price and sale year from the PPD.

Table 1. Data sources, definitions, and sample values for key variables.

Variable	Data Source	Definition	Sample Value
Number of habitable rooms	EPC	Count of habitable rooms (living rooms, dining rooms, bedrooms, studies, and similar spaces, including large kitchen-diners and non-separated conservatories). Excludes kitchens, bathrooms, utility rooms, hallways, etc.	2
Total useful floor area (m2)	EPC	Total enclosed floor area measured to internal walls (gross internal area, in square meters).	70
Current energy efficiency score	EPC	The EPC energy efficiency score on a 0–100 scale, based on the estimated annual energy cost (fuel usage for heating, hot water, lighting) per unit area. Higher = more efficient.	83
Potential energy efficiency score	EPC	The EPC potential score (0–100) that the property could achieve if recommended cost-effective improvements are implemented.	92
Current energy efficiency rating	EPC	The letter grade (A–G) corresponding to the current efficiency score (A = most efficient, G = least efficient). For example: A > 92, B = 81–91, C = 69–80, D = 55–68, etc.	B
Potential energy efficiency rating	EPC	The letter grade (A–G) corresponding to the potential efficiency score (using the same thresholds as above).	B
Transaction price (£)	PPD	Sale price of the property (at transaction date, nominal GBP).	450,000
Transaction year	PPD	Year of sale (derived from the transaction date).	2014
Transaction month	PPD	Month of sale (derived from the transaction date).	10
Property type	PPD	The property type code indicating the structural form of the dwelling: F = Flat/Maisonette, D = Detached, S = Semi-Detached, T = Terraced.	D
Borough	PPD	The borough in which the property is located, based on administrative boundaries.	Kensington and Chelsea
Energy efficiency gap	Derived	Difference between the potential and current energy efficiency scores (i.e., unrealized efficiency points).	12

summarizes the data sources and definitions of the key variables.

Our analysis is restricted to transactions from 2013 onward. This cutoff is motivated by a pivotal regulatory change in 2012 that materially affected EPC coverage and its visibility in the housing market. The Energy Performance of Buildings (England and Wales) Regulations 2012 (effective 9 January 2013) mandated for the first time that a property’s EPC rating be included in all advertising and listings at the time of sale or rent. In practice, from 2013 onward, prospective buyers were unambiguously informed of a home’s EPC rating during the marketing stage. This policy change represents an “information shock” that greatly increased the salience of energy efficiency in housing decisions [21]. By focusing on the post-2012 period, we hold constant the informational framework around energy efficiency. Any price effects we observe can thus be more confidently attributed to buyers’ valuation of energy performance rather than to changing disclosure rules.

We limit our sample to Greater London to control for regional heterogeneity and to minimize confounding from broader geographic differences across the UK (Figure 1). London offers a large, dense set of transactions under a relatively uniform economic and regulatory context, allowing for cleaner like-for-like comparisons. This focus means that our results are conditional on the London context and may not directly generalize to other regions. However, in principle, our methodology could be readily applied to other regions to explore variations in the valuation of energy efficiency.

Since the PPD data do not contain a unique property identifier (such as a UPRN) to directly link transactions across datasets, we implement a rigorous address-matching procedure to merge the PPD and EPC records. We restrict matching to exact or near-exact address pairs and impose an additional condition: the EPC inspection date must occur shortly before the transaction date and after the previous sale to ensure that the EPC data reflect the state of the property at the time of sale. This cautious matching strategy enhances data reliability but limits the sample size. Ultimately, we identify approximately 16k valid repeat sales in Greater London for which both high-quality transaction data and energy performance data are available.

Within this matched dataset, transaction volumes average around 1500 per year, with a notable dip in 2016–2018. The property-type composition remains broadly stable over time, with detached houses accounting for about 8.6% of transactions, flats for 27.2%, semi-detached houses for 24.1%, and terraced houses for 40.1%. Median prices climb steadily from about £340,000 in 2013 to just over £570,000 in 2021, closely tracking the official London average price trend (Figure 2). Because our dataset represents only a subset of the wider London market, we regard the median as a more representative and less biased measure than the mean. Across the distribution, prices rose broadly from 2013 to 2021: P10 £195k→£315k (+62%), P25 £248k→£403.8k (+63%), median £340k→£570k (+68%), P75 £524k→£843k (+61%), P90 £835k→£1.35m (+62%). P90/P10 ratio (4.28→4.29) is essentially flat, indicating broad-based appreciation rather than tail-driven shifts. Average energy efficiency scores improve by 20.3%, rising from 54 to 65, while potential energy efficiency scores remain largely unchanged, likely reflecting incremental upgrades to the existing housing stock rather than large-scale retrofits (Figure 3). The full composition table is provided in Appendix A 

Table A1. Year-by-Year Sample Composition and Summary Statistics.

Metric	2013	2014	2015	2016	2017	2018	2019	2020	2021
Transactions (count)	3105	2381	1830	1428	1363	1272	1575	1926	1503
Total Volume (£)	1.48 × 109	1.27 × 109	1.08 × 109	9 × 108	9.24 × 108	8.5 × 108	1.18 × 109	1.46 × 109	1.16 × 109
Avg Price (£)	477,594.8	531,925.6	590,433.6	630,157.3	677,983.4	668,314.8	748,864.8	756,918.5	774,908.6
Median Price (£)	339,950	390,000	451,755	500,000	538,000	535,000	560,000	590,000	570,000
Price SD (£)	520,719.1	554,555.1	485,139.2	434,919	562,285.7	480,114.7	715,964.7	632,340.6	796,041.4
P10 Price (£)	195,000	220,000	260,000	295,700	315,000	310,000	320,000	337,000	315,000
P25 Price (£)	248,000	284,000	325,000	365,000	395,000	386,000	405,000	425,000	403,750
P50 Price (£)	339,950	390,000	451,755	500,000	538,000	535,000	560,000	590,000	570,000
P75 Price (£)	524,000	595,000	670,000	740,000	755,000	760,625	845,000	850,000	843,000
P90 Price (£)	835,000	930,000	1,055,750	1,100,000	1,190,000	1,175,000	1,295,000	1,300,000	1,350,000
Mean Price Per Unit (£/m2)	4904.693	5552.447	5935.583	6284.202	6417.528	6287.13	6471.953	6727.176	6907.218
Median Price Per Unit (£/m2)	4241.071	4788.136	5248.239	5714.286	5774.648	5844.757	5824.176	6170.012	6283.784
Current EE Score (score)	54.13366	55.0651	55.12951	56.743	58.54072	62.29245	64.44444	65.46366	65.8024
Potential EE Score (0–100)	77.02963	78.17682	79.39836	79.90196	79.94571	80.18082	80.65841	80.64486	80.13307
Avg Rooms (count)	4.458615	4.440151	4.610383	4.597339	4.735877	4.709906	4.84254	4.790758	4.638723
Median Rooms (count)	4	4	5	4	5	5	5	5	4
Avg Area (m2)	92.39929	92.1884	97.45342	99.28901	104.3183	105.5987	111.8255	110.1423	107.0971
Median Area (m2)	83	83	87	87	91	92	99	100	90
Ptype D Share (%)	7.697262	8.315834	8.251366	9.593838	8.657373	8.962264	8.761905	8.15161	8.582834
Ptype F Share (%)	28.438	29.56741	25.68306	25.21008	24.6515	25.78616	23.04762	26.68744	35.46241
Ptype S Share (%)	22.83414	22.76354	25.19126	24.57983	26.85253	26.10063	24.88889	23.57217	20.42582
Ptype T Share (%)	41.0306	39.35321	40.87432	40.61625	39.83859	39.15094	43.30159	41.58879	35.52894
Age England And Wales 1900–1929 Share (%)	28.82448	27.76144	28.19672	30.88235	27.73294	29.16667	31.11111	32.39875	31.20426
Age England And Wales 1930–1949 Share (%)	28.21256	28.47543	30.16393	28.08123	33.30888	30.58176	31.04762	28.92004	26.67997
Age England And Wales 1950–1966 Share (%)	8.824477	8.609828	9.945355	9.943978	11.0785	10.61321	7.809524	8.982347	8.715902
Age England And Wales 1967–1975 Share (%)	5.120773	5.333893	5.027322	6.442577	4.328687	5.188679	4.888889	5.192108	6.254158
Age England And Wales 1976–1982 Share (%)	2.254428	2.225955	1.584699	1.540616	1.980924	1.965409	2.095238	2.336449	2.661344
Age England And Wales 1983–1990 Share (%)	3.220612	3.653927	3.114754	3.081232	2.934703	3.301887	3.428571	2.803738	4.25815
Age England And Wales 1991–1995 Share (%)	1.15942	1.637967	1.36612	0.910364	1.247249	1.493711	1.460317	1.557632	2.39521
Age England And Wales 1996–2002 Share (%)	1.674718	1.805964	1.256831	1.330532	2.054292	1.493711	2.222222	1.505711	2.59481
Age England And Wales 2003–2006 Share (%)	1.223833	1.259975	0.983607	0.70028	0.880411	1.022013	1.206349	0.674974	1.596806
Age England And Wales 2007–2011 Share (%)	0.57971	0.671987	0.437158	0.35014	0.513573	0.157233	0.126984	0.778816	0.665336
Age England And Wales Before 1900 Share (%)	18.90499	18.56363	17.9235	16.73669	13.93984	15.01572	14.60317	14.84943	12.97405
Month 01 (% of transactions)	7.149758	9.533809	8.63388	9.383754	6.236244	7.704403	6.793651	6.490135	15.03659
Month 02 (% of transactions)	6.795491	8.98782	6.994536	7.492997	6.896552	7.54717	6.349206	5.399792	14.23819
Month 03 (% of transactions)	6.859903	8.441831	7.26776	15.19608	7.923698	7.54717	6.730159	6.542056	19.76048
Month 04 (% of transactions)	7.665056	8.48383	8.087432	6.232493	6.749817	6.996855	6.920635	3.686397	7.850965
Month 05 (% of transactions)	8.792271	8.147837	8.032787	6.862745	7.410125	6.446541	7.047619	3.167186	5.256154
Month 06 (% of transactions)	7.665056	8.357833	9.180328	9.243697	9.170946	8.569182	7.428571	5.451713	16.96607
Month 07 (% of transactions)	9.243156	9.071819	11.31148	8.543417	9.757887	8.883648	10.47619	7.632399	1.929474
Month 08 (% of transactions)	11.01449	8.609828	9.453552	7.983193	9.97799	10.84906	10.22222	8.047767	3.592814
Month 09 (% of transactions)	8.373591	7.391852	7.704918	7.633053	10.93177	8.726415	9.333333	10.85151	6.786427
Month 10 (% of transactions)	8.856683	8.651827	8.961749	7.352941	7.997065	8.962264	10.4127	13.75909	2.927478
Month 11 (% of transactions)	9.822866	7.139857	6.939891	6.932773	9.170946	10.14151	9.396825	14.07061	2.59481
Month 12 (% of transactions)	7.761675	7.181856	7.431694	7.142857	7.776963	7.625786	8.888889	14.90135	3.060546
Borough Bromley Share (%)	7.536232	8.567829	7.978142	7.703081	9.904622	8.726415	8.698413	8.203531	7.451763
Dorough Wandsworth Share (%)	7.181965	5.543889	5.191257	5.812325	4.915627	6.053459	5.396825	6.697819	6.320692
Dorough Havering Share (%)	5.217391	6.257875	5.519126	6.582633	6.969919	6.289308	6.285714	6.022845	4.99002
Dorough Sutton Share (%)	5.507246	5.459891	5.68306	6.232493	6.896552	5.974843	6.285714	5.867082	5.655356
Dorough Croydon Share (%)	5.47504	5.921882	6.284153	6.302521	4.988995	6.053459	4.825397	5.192108	5.189621
Dorough Richmond Upon Thames Share (%)	5.507246	4.535909	5.519126	4.481793	4.548789	4.323899	4.825397	4.932503	4.25815
Dorough Barnet Share (%)	4.541063	5.039899	4.590164	5.112045	4.84226	4.08805	4.698413	3.478712	4.657352
Dorough Kingston Upon Thames Share (%)	3.671498	4.745905	4.754098	5.252101	4.548789	4.559748	4.31746	4.361371	4.99002
Dorough Lewisham Share (%)	4.154589	4.157917	4.043716	3.571429	3.961849	3.852201	4.571429	5.244029	4.99002
Dorough Merton Share (%)	3.89694	4.283914	4.644809	5.042017	4.108584	3.066038	3.746032	4.413292	4.25815
Dorough Other Share (%)	47.31079	45.48509	45.79235	43.90756	44.31401	47.01258	46.34921	45.58671	47.23886

.

While the dataset captures a broad range of transactions and property characteristics, it represents only a subset of the London housing market. This coverage limitation may introduce selection bias. Moreover, energy performance data are sourced from EPC certificates, which can exhibit measurement inconsistencies, particularly in older assessments, potentially introducing noise into the estimation of the green premium [32,44,45].

4. Results

4.1. Model and Result

This Section presents the results of the first-difference regression model. The analysis uses 8000 repeat-sale pairs, representing properties with an average of 2.0 sales each. The model explains approximately 82% of the variation in within-property price changes (R² = 0.822), indicating a strong overall fit. As outlined in the methodology, the analysis focuses on three key explanatory variables:

• Energy efficiency score—the change in actual EPC rating (1–100 scale) between two repeat sales.
• Energy efficiency gap—the change in the difference between potential and current EPC scores, capturing unrealized efficiency potential.
• Interaction terms between these variables and year dummies—to capture time-varying effects in their impacts.

Table 2. (A). Year-specific effects of energy efficiency (EE) score and EE gap on within-property price changes, baseline repeat-sales model (Greater London, 2013–2021; dependent variable: Δlog price). (B). Time-fixed effects (year and month) from the baseline repeat-sales model (Greater London, 2013–2021; dependent variable: Δlog price). Estimates are relative to the 2013 base year and January base month. (C). Structural characteristics and property-type interactions with energy efficiency (EE) score and EE gap in the baseline repeat-sales model (Greater London, 2013–2021; dependent variable: Δlog price).

Variable	Coefficient	Std. Err.	t-Stat	p-Value	Signif.
(A)
Energy Efficiency Score (base)	–0.0008	0.001	–1.451	0.147
Year 2014 × EE Score	0.0003	0.001	0.639	0.523
Year 2015 × EE Score	0.0021	0.001	3.204	0.001	***
Year 2016 × EE Score	0.0027	0.001	3.557	0.000	***
Year 2017 × EE Score	0.0042	0.001	5.263	0.000	***
Year 2018 × EE Score	0.0031	0.001	3.317	0.001	***
Year 2019 × EE Score	0.0041	0.001	4.710	0.000	***
Year 2020 × EE Score	0.0043	0.001	5.033	0.000	***
Year 2021 × EE Score	0.0053	0.001	5.928	0.000	***
Energy Efficiency Gap (base)	–0.0040	0.001	–6.925	0.000	***
Year 2014 × EE Gap	0.0001	0.001	0.260	0.795
Year 2015 × EE Gap	0.0007	0.001	1.027	0.305
Year 2016 × EE Gap	0.0012	0.001	1.510	0.131
Year 2017 × EE Gap	0.0025	0.001	2.978	0.003	***
Year 2018 × EE Gap	0.0019	0.001	2.002	0.045	**
Year 2019 × EE Gap	0.0025	0.001	2.767	0.006	***
Year 2020 × EE Gap	0.0047	0.001	5.207	0.000	***
Year 2021 × EE Gap	0.0069	0.001	7.380	0.000	***
(B)
Year 2014	0.1490	0.040	3.760	0.000	***
Year 2015	0.1459	0.049	2.953	0.003	***
Year 2016	0.2048	0.060	3.440	0.001	***
Year 2017	0.1176	0.063	1.868	0.062	*
Year 2018	0.1900	0.072	2.624	0.009	***
Year 2019	0.1054	0.068	1.549	0.121
Year 2020	0.0847	0.067	1.274	0.203
Year 2021	0.0157	0.069	0.226	0.821
February	–0.0089	0.007	–1.287	0.198
March	–0.0128	0.007	–1.940	0.052	*
April	0.0038	0.007	0.529	0.597
May	0.0132	0.007	1.863	0.062	*
June	0.0280	0.007	4.198	0.000	***
July	0.0387	0.007	5.753	0.000	***
August	0.0372	0.007	5.567	0.000	***
September	0.0544	0.007	8.050	0.000	***
October	0.0412	0.007	6.106	0.000	***
November	0.0559	0.007	8.254	0.000	***
December	0.0574	0.007	8.336	0.000	***
(C)
Floor Area (sqm)	0.0022	0.00008	27.299	0.000	***
Number of Rooms	0.0178	0.002	8.977	0.000	***
Flat × EE Score	0.0002	0.001	0.318	0.751
Semi-detached × EE Score	0.0014	0.001	2.848	0.004	***
Terraced × EE Score	0.0010	0.000	1.994	0.046	**
Flat × EE Gap	–0.0001	0.001	–0.166	0.868
Semi-detached × EE Gap	0.0004	0.001	0.780	0.435
Terraced × EE Gap	0.0002	0.001	0.415	0.678

The asterisks in the table indicate significance levels: * p < 0.1, ** p < 0.05, and *** p < 0.01.

A presents the estimated effects of the key energy-related variables. The baseline coefficient for the energy efficiency score (−0.0008) is negative but statistically insignificant, suggesting no clear price effect in the reference year (2013), while the coefficient for the energy efficiency gap (−0.0040, p < 0.001) is negative and highly significant, implying a price discount for properties with larger efficiency shortfalls in the reference year. The interaction between energy efficiency score and year becomes statistically significant from 2015 onward, with coefficients rising from 0.0021 to 0.0053 by 2021, indicating a growing premium for energy performance improvements. Similarly, the interaction between energy efficiency gap and year becomes significant starting in 2017 and increases to 0.0069 by 2021, suggesting that unrealized efficiency potential has also become more salient in price formation over time.

The temporal fixed effects (Table 2B) rise sharply from 2014 to a peak of 0.2048 in 2016, then decline steadily through 2021, with slower growth or contraction in 2017–2020 and near-zero effect in 2021. This trajectory mirrors the London housing percentage change index, which also shows strong gains in 2014–2016 followed by a slowdown. The main discrepancies are in 2018, when our model estimates a 0.1900 (p < 0.01) effect despite a −0.49% London price change. Differences are likely driven by variations in sample composition. Month-fixed effects reveal clear seasonal patterns: Spring, summer and early autumn months show positive effects, ranging from 0.028 to 0.0544, while February and March are lower and less significant. These seasonal patterns are consistent with prior findings in the UK housing market [46]. The coefficient is also high in December, which may be due to the sample selection bias.

Table 2C presents the results for the physical characteristic controls. Both structural controls are highly significant (p < 0.01): each additional square meter adds roughly 0.22% to price, and each extra room adds about 1.78%. In terms of property types, the energy efficiency score has a stronger positive effect for semi-detached (0.0014, p < 0.01) and terraced homes (0.0010, p < 0.05), but no significant effect for flats. This may reflect the greater scope for energy performance to influence operating costs and perceived comfort in larger, non-apartment housing forms, where buyers can more directly benefit from efficiency improvements. However, interaction terms between the energy efficiency gap and property types are not statistically significant, suggesting limited variation in how unrealized efficiency potential is valued across housing forms.

4.2. Robustness Check Result

To assess the robustness of our findings, and to ensure that the estimated temporal green premium is not driven by specific modeling assumptions or idiosyncrasies in the data, we conducted a series of sensitivity analyses. Each alternative specification tests a different potential source of bias or mismeasurement while preserving the overall repeat-sales framework.

We conducted four main sensitivity analyses:

(a)

Restricting the sample to repeat sales within three years: In this specification, we only include pairs of transactions that occurred within roughly three years of each other. This tests whether our results are influenced by very long holding periods, during which macroeconomic shifts, major renovations, or neighborhood changes could occur. Focusing on shorter intervals between sales helps isolate the price effect of energy efficiency improvements by minimizing confounding from long-horizon factors.

(b)

Using categorical EPC ratings instead of numeric scores: Our baseline model uses the numeric EPC score (1–100). However, buyers are often more familiar with the letter grade (A–G). In this alternative model, we use the EPC rating bands as the measure of energy performance. This tests whether our observed effects are robust to a coarser but more salient representation of energy efficiency.

(c)

Excluding transactions with no change in energy efficiency: In the repeat-sales sample, some properties show no change in their EPC score between sales. These cases might reflect measurement noise or truly no efficiency improvements. We remove these observations to ensure we focus on homes with actual performance shifts, making the estimated premium for efficiency gains more credible.

(d)

Controlling for borough-by-year fixed effects: To address potential geographic compositional shifts in the repeat-sales sample over time, we include interaction terms between borough and year indicators. This accounts for borough-specific price trends that could confound our temporal estimates if, for instance, later-year transactions disproportionately occur in gentrifying or more affluent areas.

The results are summarized in

Table 3. Coefficient estimates from robustness check models, including energy efficiency (EE) score and its year-by-year interaction terms, and EE gap and its year-by-year interaction terms. Full regression outputs are reported in Appendix B (Table A2, Table A3, Table A4 and Table A5).

Energy Efficiency Variable	Baseline Model	Within 3 Years	ΔEE Only	Rating Model	Borough–Year
energy_efficiency_score	−0.0008	−0.0010	−0.0008	-	−0.0009 *
energy_efficiency_rating	-	-	-	−0.0043
year_2014_energy_efficiency	0.0003	0.0013 *	0.0004	0.0052	0.0008
year_2015_energy_efficiency	0.0021 ***	0.0029 ***	0.0021 ***	0.0411 ***	0.0019 ***
year_2016_energy_efficiency	0.0027 ***	0.0027 ***	0.0025 ***	0.0470 ***	0.0018 **
year_2017_energy_efficiency	0.0042 ***	0.0046 ***	0.0043 ***	0.0605 ***	0.0032 ***
year_2018_energy_efficiency	0.0031 ***	0.0027 **	0.0031 ***	0.0405 ***	0.0015 *
year_2019_energy_efficiency	0.0041 ***	0.0036 ***	0.0041 ***	0.0476 ***	0.0026 ***
year_2020_energy_efficiency	0.0043 ***	0.0034 ***	0.0041 ***	0.0469 ***	0.0031 ***
year_2021_energy_efficiency	0.0053 ***	0.0050 ***	0.0054 ***	0.0487 ***	0.0036 ***
energy_efficiency_score_gap	−0.0040 ***	−0.0042 ***	−0.0040 ***	−0.0439 ***	−0.0033 ***
year_2014_energy_efficiency _gap	0.0001	0.0015 **	0.0002	0.0001	0.0006
year_2015_energy_efficiency _gap	0.0007	0.0017 **	0.0006	0.0102	0.0003
year_2016_energy_efficiency _gap	0.0012	0.0019 **	0.0009	0.0178 **	0.0001
year_2017_energy_efficiency _gap	0.0025 ***	0.0030 ***	0.0025 ***	0.0303 ***	0.0009
year_2018_energy_efficiency _gap	0.0019 **	0.0018	0.0019 **	0.0252 ***	0.0000
year_2019_energy_efficiency _gap	0.0025 ***	0.0022 *	0.0023 **	0.0357 ***	0.0004
year_2020_energy_efficiency _gap	0.0047 ***	0.0048 ***	0.0046 ***	0.0486 ***	0.0028 ***
year_2021_energy_efficiency _gap	0.0069 ***	0.0072 ***	0.0069 ***	0.0685 ***	0.0046 ***

The asterisks in the table indicate significance levels: * p < 0.1, ** p < 0.05, and *** p < 0.01.

, with coefficients and significance levels reported for key energy-efficiency-related terms across models.

5. Discussion

The results above are best understood in the context of the evolving policy landscape surrounding residential energy performance in the UK. Between 2014 and 2021, a series of government initiatives gradually elevated the importance of energy efficiency and appear to have directly influenced buyer behavior and willingness-to-pay for energy-related attributes. Below, we interpret the results considering these policy and market developments.

5.1. Growing Time-Varying Premiums

We find a strong and statistically significant time-varying premium for energy efficiency improvements, with two distinct upward-trending phases separated by a temporary setback in 2018. While the direct effect of the energy_efficiency_score is not statistically significant in our model (−0.0008, p = 0.147), the interaction terms between efficiency improvements and year dummies (year_X_energy_efficiency) show increasingly positive coefficients over time. This indicates that the timing of an efficiency upgrade plays a critical role: the same level of improvement is valued more in later years than in earlier ones. Specifically, a one-point improvement in EPC score is associated with a 0.13% (−0.0008 + 0.0021) price increase in 2015, rising to 0.19% in 2016 and 0.34% in 2017—marking the first upward-trending phase (2015–2017). In 2018, however, the coefficient falls to 0.23%, suggesting a temporary green premium setback. This pattern marks two distinct upward-trending periods, 2015 to 2017 and 2019 to 2021, separated by the 2018 dip (Figure 4).

If we trace relevant policy changes over this period, a clear pattern emerges. The 2015 Energy Efficiency (Private Rented Property) Regulations marked the start of a firmer regulatory stance, setting a minimum EPC rating of “E” for the private rented sector. Although targeted at rentals, this raised the visibility and salience of EPC ratings across the market, signaling future tightening and reinforcing energy performance as part of long-term asset value. Consistent with this shift, the first upward period (2015–2017) aligns with the initial diffusion of EPC information into pricing: buyers and valuers appear to have begun capitalizing efficiency improvements soon after 2015, with steadily larger price effects through 2017. In 2018, however, the estimated premium moderates despite MEES coming into force (April 2018). While MEES coincides with price discounts for F/G-rated homes [27], the temporary setback in the green premium is more plausibly tied to macro headwinds, notably Brexit-related uncertainty, a cooling London market with softer transaction volumes and slower nominal growth, rising borrowing costs, and composition effects (e.g., shifts in the mix of transacting properties). These forces likely dampened buyers’ willingness (or ability) to pay extra for efficiency in that year, partially offsetting policy-driven salience. The second upward period (2019–2021) then reflects a renewed and stronger capitalization of energy performance. This acceleration is consistent with the UK’s 2019 legal commitment to net-zero by 2050 and the 2020 extension of MEES to existing tenancies, which increased regulatory pressure on low-efficiency stock. At the same time, broader fundamentals amplified the pricing of efficiency: rising energy costs in the late 2010s increased the perceived operating savings from efficient homes; public climate awareness intensified; EPC data became easier to access and compare in listings, reducing information frictions. Together, these policies and market forces help explain why the green premium resumed its rise after 2018 and reached its peak by 2021.

However, the upward trend in the green premium cannot be attributed to policy alone; it also reflects shifts in market fundamentals and buyer sentiment. Rising energy costs in the late 2010s heightened attention to operating expenses and increased the perceived value of efficient homes. Public climate awareness intensified, which is reinforced by the UK’s 2019 legal commitment to net-zero by 2050, elevating sustainability in purchase decisions. At the same time, improved access to EPC data and its systematic inclusion in property listings since 2013 reduced information frictions, enabling buyers to price energy performance more consistently. The COVID-19 period further focused households on in-home quality and running costs as time spent at home rose from 2020. Taken together, these forces—distinct from but complementary to regulation—help explain the renewed acceleration in the premium during 2019–2021 and suggest that valuations increasingly favor energy-efficient homes due to both compliance considerations and expectations of tangible cost savings.

5.2. Reframing the Energy Efficiency Gap

The regression results reveal a notable temporal shift in how the market values a home’s energy efficiency gap, i.e., the unrealized efficiency potential indicated by the difference between current and possible EPC rating (energy_efficiency_gap). Initially, a larger gap carried a significant price penalty. In the baseline year (2013), each one-point increase in the efficiency gap is associated with roughly a 0.4% lower sale price (coef. –0.0040, p < 0.001), meaning less efficient homes sold at a discount, all else equal. However, this discount steadily diminished over time and eventually reversed. The yearly interaction terms are insignificant through 2016, then turn positive and significant from 2017 (coef. 0.0025, p < 0.003) onward, eroding the negative effect. By 2020 the net effect of the gap on price is just above zero (−0.0040 + 0.0047), and by 2021 it becomes positive (~+0.3% per point, −0.0040 + 0.0069) (Figure 5). The data suggest that as of 2020–2021 the market no longer treats a high efficiency gap strictly as a liability; if anything, homes with greater scope for efficiency improvements were fetching equal or higher prices by the end of the study period.

These findings indicate a major change in buyer perceptions and incentives in London’s housing market. Early in the decade, higher-rated (more efficient) properties consistently commanded price premiums, while the poorest-rated homes suffered discounts. Our 2013–2019 results align with this pattern of a “brown discount.” Buyers at that time were evidently wary of homes with poor efficiency or many upgrades left undone, pricing them lower to account for anticipated retrofit costs or inferior energy performance. However, several factors likely underpinned a change from 2020. First, in 2020 a “Green Homes Grant” scheme briefly offered subsidies for insulation, efficient heating, and other improvements. Such initiatives lowered the effective cost of upgrading an inefficient property, thereby reducing buyers’ perceived risk in purchasing homes with a large efficiency gap. In fact, some buyers might have preferred making the improvements themselves to benefit from any available subsidies or to ensure quality, rather than paying a premium for a home where upgrades were already done. They also might have wanted to capture potential value creation for themselves from the upgrades. Second, market-wide improvements in energy efficiency narrowed the gap between efficient and inefficient homes. Over the study period, the average energy rating of the properties had improved by almost 20% (Figure 2). This mass improvement means the typical “inefficient” home in 2021 was likely less egregiously inefficient than in 2013. As a result, the financial and comfort downsides of a given efficiency gap were smaller in 2021 than before, warranting a smaller price penalty. Third, in the robust housing market of 2020–2021 (fueled in part by pandemic-era dynamics such as increased savings, low interest rates, and a desire for homes with more space) buyers may have been more willing to overlook or even embrace a home’s unrealized efficiency potential. Thus, a house with a big efficiency gap (say an unrestored Victorian townhouse) might still attract bidding wars for its other attributes. Fourth, the anticipation of stricter standards (the government in 2021 was consulting on raising the minimum to EPC C for rentals by 2025–2028) meant savvy buyers knew that efficiency upgrades were a matter of when not if. An inefficient home’s price already reflected that impending upgrade cost, so additional “penalty” depreciations diminished. In short, the market moved from a “brown discount” to a largely neutral environment. A valuable next step for further research would be to test these mechanisms using buyer-level data to observe how preferences and expectations evolved at the household level.

5.3. Robustness and Sensitivity: Confirming Temporal Dynamics

We further take a closer look at the robustness check results to assess whether our findings could be influenced by model specification choices or shifts in the underlying data composition. These sensitivity checks consistently confirm the stability of our estimates, reinforcing the robustness of the observed temporal green premium.

Year-Specific Energy Efficiency Effects: Across specifications, the price response to efficiency improvements strengthens over time and consistently exhibits a two-phase pattern: steady gains from 2013 to 2017, a modest dip in 2018, and renewed growth through 2021. In the three score-based models (Baseline, 3-Year, ΔEE-only), year-specific coefficients follow almost identical trajectories, rising from about 0.002 in 2015 to above 0.004 in 2017, dipping to 0.0027–0.0031 in 2018, and climbing to 0.0050–0.0054 in 2021 (p < 0.001). The base-year coefficients are slightly negative in all three models (−0.0008 to −0.0010), so the net effects turn positive after 2014 and converge by 2021 to premiums of roughly 0.44–0.54% per point. The close alignment across these specifications, despite differences in sample composition restrictions, demonstrates that the upward temporal trend is not sensitive to resale interval or conditioning on observed improvements, confirming the robustness of the main finding.

The other two sensitivity models reinforce the main finding through both cross-metric triangulation and design-based robustness. The categorical rating model produces a substantially larger 2021 coefficient (0.0487, p < 0.001) for a one-letter EPC upgrade, which translates to an equivalent score-based premium of ≈0.0041–0.0061 per point (averagely 12 points per letter), closely matching the range from the score-based models and confirming that the magnitude difference reflects scale rather than a distinct pricing mechanism. The borough-by-year fixed-effects model, which rigorously controls for neighborhood-specific price dynamics, also preserves the two-phase trajectory, with coefficients of 0.0009 in 2013, 0.0032 in 2017, 0.0015 in 2018, and 0.0036 in 2021 (p < 0.001), yielding a smaller but still positive net premium by the end of the sample. Together, these results demonstrate that the strengthening capitalization of energy efficiency is robust to different sample restrictions, alternative functional forms, measurement scales, and spatial fixed-effects structures (Figure 6). The temporal profile—emergence, dip, and consolidation—fits a diffusion dynamic in which information salience and buyer learning increase over time. While we do not claim causality, the convergence across specifications and scales supports a durable, strengthening capitalization of energy efficiency.

Energy Efficiency Gap and Interactions: Across all model specifications, the main effect of the energy_efficiency_gap term is negative and highly significant, indicating that properties with greater unrealized efficiency potential have been consistently discounted in the housing market since 2013. Base-year coefficients range from −0.0033 in the borough-by-year fixed-effects model to −0.0439 in the categorical rating model, reflecting persistent price penalties that likely capture anticipated retrofit costs or perceived inefficiency.

The year-specific interaction terms reveal a gradual but sustained weakening of this discount over time, although the pace and magnitude vary by specification. In the 3-Year repeat-sales model, the gap’s 2021 interaction term reaches 0.0072 (p < 0.001), up from an insignificant 0.0015 in 2014, effectively offsetting and surpassing the baseline penalty by the end of the sample. The categorical rating model shows an even sharper shift: interaction terms rise from negligible levels in early years to 0.0685 in 2021 (p < 0.001), consistent with a stronger repricing effect when efficiency differences are framed categorically. The borough-by-year fixed-effects model produced a more muted but still similar temporal trajectory, with interaction terms remaining insignificant until 2020, when they turn positive and significant (0.0028 in 2020; 0.0046 in 2021, both p < 0.001). Although the magnitude of effects varies across models, the consistent emergence of positive and statistically significant interaction terms in recent years reinforces the finding that the market’s valuation penalty of retrofit potential may have decreased over time (Figure 7).

6. Conclusions

This paper fills the above-described gap by providing a systematic year-by-year analysis of the green premium in the London housing market over the period 2013–2021. In doing so, it builds on the prior literature and extends it in two critical ways.

First, we explicitly trace how the market’s pricing of energy performance evolved annually during a time of significant policy intervention and rising environmental consciousness. Rather than assuming a constant effect of EPC ratings, our hedonic framework allows the premium associated with a given energy efficiency improvement to vary by year. This approach captures the temporal dynamics that other studies have largely treated as a black box. Indeed, our findings indicate that the green premium in London rose from statistically insignificant levels in 2013 to approximately 0.47% in 2021, with a temporary dip in 2018, broadly aligning with the rollout of stronger energy-efficiency policies and growing market attention to sustainability. Buyers in recent years appear to place greater monetary value on energy-efficient dwellings than they did in the mid-2010s, a trend consistent with the tightening of regulations and with higher expected energy costs. Such evidence answers the call in recent reviews for up-to-date longitudinal analysis of EPC impacts and demonstrates that the market’s valuation of “green” attributes is not static.

Second, our study introduces a nuanced distinction between achieved energy performance and unrealized energy-saving potential in a property and examines how the market prices each of these measures over time. We decompose a home’s energy efficiency characteristics into (a) its current EPC efficiency score and (b) an “energy efficiency gap,” defined as the difference between the home’s current score and its potential score. This conceptual innovation builds on the idea that two homes with the same present efficiency might not be valued equally if one has much greater scope for easy improvement. The literature to date has mostly focused on the levels of efficiency achieved, without considering how untapped potential or the required retrofit effort factors into pricing. Our results reveal a striking dynamic: in the early years of our sample, a large efficiency gap, which means a home had significant room for improvement, was associated with a price penalty. Buyers before 2020 appeared to discount properties that were inefficient relative to their potential, presumably due to anticipated retrofit costs or a signal of neglect. However, over time this penalty diminished and by the end of the period had vanished or even turned into a slight premium. By 2020–2021 the market no longer heavily penalized homes for unrealized efficiency improvements. This temporal reversal in the valuation of the efficiency gap is a novel finding. This attenuation is consistent with a confluence of forces documented in our analysis: the 2020 Green Homes Grant temporarily lowered expected retrofit costs; stock-wide gains in energy performance (~20% increase in the average rating) narrowed the dispersion between efficient and inefficient homes; and a buoyant pandemic-era market shifted buyer focus toward other attributes and made “upgrade potential” less of a deterrent. To our knowledge, this paper is the first to document such an effect, highlighting an evolving market logic: both actual energy performance and latent potential are being priced, and their price impacts can change over time.

Overall, the present study adds a new dimension to the green premium literature. It provides empirical evidence that the relationship between energy efficiency and housing value is not static, but is responsive to regulatory, economic, and social changes. By tracking year-by-year changes in London, a leading global city with an intense policy focus on sustainability, our analysis offers insight into how quickly market preferences adjust and how policy interventions may translate into property values. This enriched understanding helps fill the gap identified in previous research, moving the field beyond the question of “Does a green premium exist?” to the deeper questions of “How the green premium evolves”. That said, focusing on London limits external validity: regions with lower energy costs, weaker policy enforcement, or different market and household structures may exhibit smaller or differently timed premiums. For example, differences in (i) energy prices and fuel mix, (ii) MEES/EPC stringency and enforcement, (iii) housing-stock age/type, (iv) climate (heating-degree days), (v) tenure mix and landlord incentives, and (vi) socio-demographics (income, education, environmental preferences) could all moderate both the level and trajectory of the green premium. Ultimately, these findings underscore the importance of temporal analysis in real estate sustainability research and position our work as a timely contribution that bridges the static empirical evidence with the dynamic nature of housing markets’ responses to the energy efficiency agenda.