Finds documents with both search terms in any word order, permitting "n" words as a maximum distance between them. Best choose between 15 and 30 (e.g. NEAR(recruit, professionals, 20)).
Finds documents with the search term in word versions or composites. The asterisk * marks whether you wish them BEFORE, BEHIND, or BEFORE and BEHIND the search term (e.g. lightweight*, *lightweight, *lightweight*).
This article delves into the critical role of building regulations in enhancing energy efficiency in Swedish multi-apartment buildings. It explores how performance-based building codes, introduced and tightened over the years, have influenced energy consumption patterns. The study utilizes a comprehensive dataset from Energy Performance Certificates (EPCs) to assess the impact of regulatory changes on energy use in buildings constructed between 2000 and 2019. Key findings reveal significant reductions in specific energy use following the implementation of stricter building codes, with notable improvements in energy efficiency. The article also highlights the methodological rigor employed, including quantile regression and structural break analysis, to identify heterogeneous regulatory effects across different energy performance levels. Additionally, it discusses the energy performance gap, where estimated energy use often differs from actual consumption, providing valuable insights for policymakers and practitioners aiming to bridge this gap and achieve greater energy savings. The analysis underscores the importance of continuous monitoring and evaluation of building regulations to ensure they effectively contribute to climate change mitigation and energy security.
AI Generated
This summary of the content was generated with the help of AI.
Abstract
Building codes currently regulate energy efficiency in newly constructed buildings in Sweden. Alongside energy declarations, performance-based regulation specifying specific energy use requirements was introduced in Sweden in 2006. The requirements have been subsequently tightened to enhance energy performance. This study estimates the impact of these requirements on energy savings in Swedish multi-apartment buildings, relying on specific energy use data from energy performance certificates (EPCs). An estimated time trend indicates greater energy efficiency at a rate of 1.57% per year for buildings with district heating and 1.09% per year for electrically heated buildings. After accounting for this trend, the results indicate that the implementation of performance-based regulation is associated with a 14.2% increase in energy efficiency for buildings with district heating and a 9.7% increase for those with electric heating. Moreover, the first tightened building codes generates an additional 2% increases in energy efficiency for district-heated buildings and an approximately 7.4% improvement for electrically heated buildings. However, there is no evidence to suggest that the second tightening of building codes have strong effects on further increasing energy efficiency. Furthermore, the effect of building codes is more substantial for buildings where actual energy use exceeds the mandated levels and modest for buildings with better energy performance. Alternatively, when studying the time trend of energy efficiency, I find a structural break with a significantly greater increase in efficiency over time during the period of regulation compared to before. In addition, findings in this study indicate evidence of the energy performance gap, where the estimated energy use from engineering models is substantially lower than the measured energy use for comparable construction.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Increased energy efficiency has become an increasingly important goal in the last decades. It is one of the three climate and energy targets in the European Union for achieving the climate goals of 2030.1 The European Union has set a new target to reduce final energy consumption by 11.7% by 2030, relative to the 2020 reference scenario (EC, 2023). The recast Energy Efficiency Directive (EED) requires Sweden to achieve cumulative end-use energy savings of 237 TWh during the period 2021–2030 (EC, 2024).
Buildings account for approximately 40% of total final energy use in the EU. Compared to energy use in the industrial and transport sectors, the building sector has demonstrated relatively large energy-saving potential (EC, 2018). Energy efficiency in buildings refers to using less energy to heat and cool buildings and to run appliances (Patterson, 1996). More efficient buildings require less energy, making energy efficiency in buildings a cost-effective way to reduce greenhouse gas emissions and contribute to climate change mitigation. Additionally, decreased demand for energy would substantially improve energy supply security (IEA, 2019).
Advertisement
As a regulatory instrument, building regulations can increase the energy efficiency of new buildings. Newly constructed buildings are subject to technical requirements and environmental goals. Energy declarations, also known as Energy Performance Certificates (EPCs), were first mandated in the EU following the Energy Performance of Buildings Directive in 2002, which aims to support the achievement of energy efficiency targets (EC, 2002).
Together with energy declarations, Sweden has implemented performance-based building regulations since July 2006 (BFS, 2006:12 – BBR12), addressing the maximum permitted limits for unit energy use (expressed in \(\text{kWh}/{\text{m}}^{2}\text{ per year})\). Furthermore, energy requirements for the specific energy use of both electrically and non-electrically heated buildings have been subsequently tightened to enhance the energy performance of new construction (BFS, 2008:20 – BBR16; BFS, 2011:26 – BBR19; BFS, 2015:3 – BBR22). This raises an essential question: Have these changes in building codes been effective in increasing energy efficiency?
The study aims to assess the impact of the introduced and tightened performance-based building codes on energy efficiency in Swedish multi-apartment buildings. As of 2023, approximately 50% of the households live in multi-apartment buildings in Sweden, which are responsible for 36% of total energy use (SCB, 2025; SEA, 2025). Using data from EPCs, this study focuses on multi-apartment buildings constructed between 2000 and 2019.
Several studies have examined the impact of building regulations or building codes on energy use. The literature has primarily focused on empirical studies in the United States (e.g., Kotchen, 2017; Levinson, 2016; Novan et al., 2022). These studies typically compare energy use in buildings constructed before and after the implementation of building code changes. Overall, most empirical studies suggest that energy use decreases with the implementation of building codes, indicating that building energy codes have a positive effect on improving energy efficiency.
Advertisement
Comparing to previous studies, this study makes several contributions to the literature on evaluating the energy savings from building codes. First, the regulatory effects in the Swedish building sector have not been investigated using regression methods. Sweden is the only country in the Nordic region where the absolute energy use in buildings has fallen, by nearly 1% per year (Tammiste et al., 2018). It is imperative to understand the extent to which building regulations contribute to improved energy efficiency.
Second, while most previous studies use housing-level data based on surveys, the energy use data analyzed in this study comes from energy bills. In Sweden, the majority of EPCs are based on energy bills, making this database a leading source for studying energy use in the Swedish building stock. Unlike previous studies that mainly have focused on building regulation effects on total energy consumption or per capita energy consumption, specific energy use (expressed in \(\text{kWh}/{\text{m}}^{2}\text{ per year}\)) is analyzed in this study. Energy savings in terms of specific energy use can be directly compared with the requirements in the building codes, making it more suitable for capturing energy efficiency in the building sector (Odyssee-Mure, 2020).
Third, in comparison to recent empirical studies estimating the average effects of building codes, this study also makes methodological contributions by applying quantile regression to identify the heterogeneous regulatory effects across the energy performance levels of buildings. It also employs structural break analysis to examine the change in the time trend following the introduction of performance-based building codes. Several empirical challenges are also considered in this study, including policy endogeneity, unobserved trends in average energy use based on years of construction, and the gap between building permits and construction completion.
Building regulations establish minimum standards for the energy efficiency of newly constructed buildings, playing a crucial role in regulating construction quality of the building.2 Consequently, improvements in building quality result in an increase in energy efficiency associated with building characteristics.
Table 1 provides a summary of empirical studies in the residential sector that estimated the effects of building regulations or building codes on energy use. Most previous studies are from the United States, where energy consumption for electricity and natural gas in residential buildings is considered. Almost all of these studies suggest that the effects on reducing energy use for natural gas are relatively higher than the effects on reducing electricity use (e.g., Jacobsen & Kotchen, 2013; Kotchen, 2017; Levinson, 2016). However, one recent study finds that building codes increased natural gas consumption due to broader access to natural gas and its increased use in major end-use consumption categories in houses (Kellogg & Cumbre-Gibbs, 2023). In addition, one study from Mexico found finds that houses built to a higher building standard do not consume less electricity.
Table 1
Summary of selected studies on the effects of building regulations or building codes on energy use in residential buildings
aThe countries include Austria, Denmark, Finland, France, Germany, Poland, and the United Kingdom
bFirst, the general regression of energy use on year of construction, residential demographics, and building characteristics; Second, focusing on the sensitivity of energy consumption with respect to outdoor temperature changes; Third, comparing energy use in California with other states in the U.S
cThe reduction is in electricity expenditures rather than consumption
A limited number of studies have assessed the effects of building regulations or building codes on energy use in the EU. Saussay et al. (2012) found a significant impact of building energy codes on the improvement of residential space heating energy efficiency in seven European countries. Building energy codes were measured by the number of years elapsed since their enactment in each country. In recent years, EPCs have become key tools for evaluating energy consumption in the EU (Pagliaro et al., 2021; Pasichnyi et al., 2019). In Sweden, Hjortling et al. (2017) used Swedish EPC data for multi-dwelling and commercial buildings, finding that energy consumption was lower when stricter building codes were implemented. However, the consumption levels were likely to be higher than the building code requirements.
Nevertheless, Hjortling et al. (2017) primarily relied on descriptive statistics rather than regression analysis. This study fills a gap in the literature by comparing energy use between buildings constructed before and after a building code change in Swedish multi-apartment buildings, using various econometrics approaches and controlling for other factors influencing energy consumption.
Background on energy declarations and building regulations in Sweden
Energy performance certificates (EPCs) in Sweden
In Sweden, EPCs became obligatory in 2006 (Sveriges Riksdag, 2006a, b). Since then, Boverket, the Swedish National Board of Housing, Building and Planning, has overseen the regulation of energy declarations for buildings (BFS, 2007:4 -BED 1), with the regulation being amended for greater strictness in its latest version, implemented in 2021 (BFS, 2011:26 -BED 11).
EPCs are required for newly constructed buildings, those that are sold or rented out, or those with a total useful floor area exceeding 250 square meters and occupied by a public authority. When a building is renovated, the altered part must comply with the regulations for new buildings (Boverket, 2025). In Sweden, all multi-apartment buildings were required to have a registered EPC no later than the end of 2008. As the EPC is valid for 10 years, many owners of multi-apartment buildings have now obtained their second EPC (von Platten et al., 2019).
An EPC provides information on energy consumption of a building, serving as a basis for comparing the energy efficiency of different buildings. Energy performance data used typically includes information on energy consumption for heating, hot water, comfort cooling, and the building’s property electricity.3 The electricity consumption of appliances and user equipment is not included in the energy performance measure.
One important feature of EPCs in Sweden is that energy use data are derived from energy bills, providing actual energy consumption data. The building owner supplies information on the building's energy use based on energy bills to an independent expert. An inspection is then carried out, during which the energy expert evaluates the specific building, verifies the information, and issues the EPC. The energy expert is responsible for ensuring that the EPC is prepared correctly (BFS, 2007:4—BED 1). In cases energy use cannot be directly metered, such as for newly constructed buildings, experts can calculate the energy use based on the technical attributes of the buildings. However, the majority of EPCs are based on measured energy use, ensuring the accuracy of the data.
Furthermore, the energy use values in the EPCs undergo a two-step correction process. First, the data is adjusted to a standard level of energy use by accounting for factors such as domestic hot water usage, maintaining a normal indoor temperature of 21 degrees Celsius, and considering internal loads, which control for the general increase in indoor temperatures caused by thermal comfort (Mavrogianni et al., 2013). The second step involves a normal year correction, which takes into account the local normal climate conditions and the climate during the period when energy use was measured (Boverket, 2017, 2022).4
Building regulations in Sweden
Boverket’s Building regulations (BBR) were initially implemented in Sweden in 1994. Building regulations traditionally focused on requirements for each individual building component (e.g., walls, windows, roof, floor), measured in U-values (\(\text{W}/({\text{m}}^{2}\cdot \text{K})\)), to prevent heat transfer between the inside and the outside of a building (BFS, 1993:57—BBR 1). Over time, several amendments have been made in the subsequent years.5
To comply with the energy declarations and support policy goals at reducing the climate impact of buildings, performance-based regulations were introduced in 2006 (BFS, 2006:12 – BBR12). The new regulations address the overall quality of a building, setting the maximum permitted limits for unit energy use (expressed in \(\text{kWh}/{\text{m}}^{2}\text{ per year})\). In 2011, the Swedish government specified the requirements for buildings in the Planning and Building Ordinance (Sveriges Riksdag, 2011), with new regulations containing mandatory provisions (BFS, 2011:6 – BBR18). Among all the sections in the building regulation, Sect. 9 specifies the requirements for energy conservation. Dwellings and premises must be designed to ensure that the building’s final specific energy use does not exceed the regulated values. Voluntary higher efficiencies are allowed, permitting buildings that exceed the BBR requirements.
Table 2 summarizes the BBR requirements for specific energy use in residential multi-apartment buildings. The initial specific energy use requirements in BFS, 2006:12 – BBR12 were 130 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) for dwellings in the northern climate zone and 110 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) for dwellings in the southern climate zone. There were no stricter requirement for multi-apartment dwellings with electric heating compared to those mainly with district heating.
Table 2
Summary of BBR requirements on specific energy use (\(\text{kWh}/{\text{m}}^{2}\text{ per year}\)) for residential multi-apartment buildings
BBR 12
(BFS 2006–12)
BBR 16
(BFS 2008–20)
BBR 19
(BFS 2011–26)
BBR 22
(BFS 2015–3)
Dwellings that mainly have district heatinga
Climate zone I
130
150
130
115
Climate zone II
130
130
110
100
Climate zone III
110
110
90
80
Climate zone IV
75
Effective time
2006–07
(2007–07)
2012–01
(2013–01)
2015–03
(2016–03)
Dwellings that have electric heating
Climate zone I
130
110
95
-
85
Climate zone II
130
75
-
65
Climate zone III
110
55
-
50
Climate zone IV
45
Effective time
2006–07
(2007–07)
2009–02
(2010–01)
2015–03
(2016–03)
aFor dwellings with non-electric heating, the heating methods include district heating and fuels, with more than 95% using district heating.
The energy requirements for buildings with electric heating were tightened in early 2009 (BFS, 2008:20 – BBR16).6 The specific energy use requirements were set at 95 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\), 75 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\), and 55 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) for climate zones I, II, and III, respectively. With the northern climate zone divided into climate zones I and II, the required value \(\text{year}\) for dwelling mainly with district heating in climate zone I was released to 150 \(\text{kWh}/{\text{m}}^{2}\text{ per}\) year. However, the energy requirements were tightened further at the beginning of 2012 (BFS, 2011:26 – BBR19).7 The amendment set stricter requirements specific energy use requirements, reducing the values by 20 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in each climate zone for buildings mainly with district heating.
In 2015, under BFS, 2015:3 – BBR22, the energy use requirements were once again tightened for buildings using both types of energy. In connection with these tightened requirements, a new climate zone (climate zone IV) was introduced, which includes municipalities in southern and western Sweden, separating it from climate zone III (as detailed in Appendix 1). For buildings mainly with district heating, in comparison to the previous regulation (BFS, 2011:26 – BBR19), the energy use requirements were tightened by 10 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in climate zones II and III and by 15 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in climate zones I and IV. For electrically heated dwellings, relative to the previous regulation, requirements were tightened by 5 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in climate zone III and by 10 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in climates zones I, II, and IV.
Since July 2006, Sweden has implemented performance-based building regulations. Energy requirements were subsequently tightened to enhance the energy performance of new construction. With the adoption of these new building codes, the hypothesis is that energy consumption in buildings constructed under these codes should be lower compared to buildings built under earlier regulations.
Data and empirical strategies
Sample selection
EPCs are registered in the Swedish national database GRIPEN, which is provided by Boverket (the National Board of Housing, Building and Planning). GRIPEN provides public access to limited EPC data, but the data can be fully accessed for research purposes through an established agreement.
The database includes all building categories. Residential multi-apartment buildings are selected based on the building type code. After dropping duplicated certificates using the declaration ID, the extracted dataset contains a total of 101,153 energy declarations for Swedish multi-apartment buildings. Since many building owners have obtained a second EPC, the majority of the EPCs in the dataset were approved from 2018 to 2021.
Each energy performance certificate provides information on energy use at the building level. As mentioned in the background, energy performance is measured in different ways, including specific energy use (for EPCs approved before 2019) and primary energy (for EPCs approved since 2019). However, to ensure a fair comparison of energy use, it is necessary to use a consistent measure across the dataset. Therefore, in this study, specific energy use (final energy use) from all EPCs is selected as the dependent variable.
Among all building characteristics, the key variable is the year of new construction, as it determines the stringency of the building codes which the construction faced. To illustrate the energy use trend over construction years, Fig. 1 displays the average specific energy use for multi-apartment buildings constructed since 1980. Within each year of new construction, electrically heated apartment buildings tend to use less heating energy compared to those mainly using district heating.
Fig. 1
Average specific energy use in residential multi-apartment buildings over construction years. Notes: The summary for specific energy use for buildings with non-electric heating are based on 26,646 energy performance certificates (95.61% using district heating); the summary statistics for buildings with electric heating are based on 10,450 certificates
The figure depicts a decreasing trend in specific energy use for both types of energy. However, there is a downward trend in energy use across years of construction before and after the implementation of the performance-based building codes in 2006. To distinguish the effects due to the change in building codes from other factors at the time of housing construction, it is necessary to consider the unobserved trend in average energy use based on years of construction.
As can be seen in the figure, energy use does not always have a decreasing trend. For instance, the average energy use in buildings with district heating constructed in 2000–2005 is similar to energy use in buildings constructed in the 1980 s. One possible explanation for this is that older buildings may have undergone renovations, such as improvements to windows, walls, and other features. Therefore, when measuring the actual savings caused by changes in building codes, a direct comparison for energy use between new and old buildings may not be entirely fair. One strategy for dealing with this is to restrict attention to energy use in relatively new buildings, selecting those constructed after 2000 as the reference level, to minimize the possibility that other policy instruments (such as renovation programs) will bias the analysis.
In addition, another empirical challenge arises from the lag between a change in regulation and new construction conforming to the regulation. The variable year-of-construction in the data denotes the year when construction was completed, specifically indicating when the majority of the building could be put into use. However, building codes are typically enforced when building permits are issued, which occurs before construction work starts. Therefore, when using years of construction to determine which building codes the construction work faced, the permit-to-completion period should be considered. Given that the average time to construct a new multi-apartment building is 18–24 months, the year following changes in building codes can be treated as a transition period. It becomes challenging to identify buildings constructed during this period as subject to requirements before or after the code change.
Figure 2 illustrates the basic study design and the sample selection for measuring the effects of introduced and tightened performance-based building codes on energy use. For buildings with a heating method other than electric heating (see Fig. 2a), as the performance-based regulation came into force after July 2006 (Reform 1), I exclude buildings constructed in 2007–2008 from the analysis.8 Similarly, for studying the effect of the tightened building codes in January 2012 and March 2015 (Reform 2 and Reform 3), I exclude buildings constructed in 2013 and 2016, respectively.
Fig. 2
Sample selection (a): For buildings mainly with district heating (b) For buildings with electric heating
The sample selection for electrically heated multi-apartment buildings is shown in Fig. 2b. When measuring the effects of the performance-based regulation (Reform 1), one difference compared to in Fig. 2a is that only buildings constructed in 2007 were excluded from the analysis. On the one hand, buildings with electric heating are typically smaller than those with district heating, and therefore, they are expected to have a shorter construction period.9 Buildings constructed in 2008 are more likely to be considered as built after the introduction of performance-based building codes. On the other hand, from a technical perspective, excluding two years of observations may result in too few observations remaining in the Reform 1 group. Furthermore, to study the effect of the tightened building code in 2009 and 2015 (Reform 2 and Reform 3), buildings constructed in 2010 and 2016 were separately excluded from the analysis.
In addition to years of construction indicators, other factors that may influence energy use are also taken into consideration. The total space of the building, measured by the heated area, is selected from the ECP database. The average size of apartments is then derived based on the total space and the number of apartments in the building. Furthermore, energy prices and income at the municipal level are also controlled for this study. The prices of district heating and electricity are collected from the Nils Holgersson Report (NHR, 1996), while average household income data is taken from Statistics Sweden.Additionally, given that building regulations may not be entirely harmonized at the national level, I create a dummy variable for each municipality to account for other potential unmeasured factors. Nonetheless, since specific energy use values from the EPC are already climate-corrected, the analysis does not require further controls for weather conditions.
Descriptive statistics
Table 3 summarizes the means and standard deviations for relevant variables for multi-apartment buildings, mainly for those with district heating and those with electric heating. The statistics are presented separately for buildings in the baseline group (buildings constructed between 2000 and 2006 for both types), the Reform 1 group (buildings constructed in 2009–2012 for district heating, and in 2008–2009 for electric heating), the Reform 2 group (buildings constructed in 2014–2015 for district heating, and in 2011–2015 for electric heating), and the Reform 3 group (buildings constructed between 2017 and 2019 for both energy types).
Table 3
Descriptive statistics by group of construction years
Buildings that mainly have district heating
Buildings that have electric heating
Baseline group
(2000–2006)
Reform 1 group
(2009–2012)
Reform 2 group
(2014–2015)
Reform 3 group
(2017–2019)
Baseline group
(2000–2006)
Reform 1 group
(2008–2009)
Reform 2 group
(2011–2015)
Reform 3 group
(2017–2019)
Variable
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Mean
(Std. dev.)
Panel A: Energy consumption
Specific energy use \((\mathbf{k}\mathbf{W}\mathbf{h}/{\mathbf{m}}^{2}\) per year)
126.34
(31.43)
95.52
(25.54)
86.34
(21.11)
79.10
(24.01)
80.93
(34.31)
68.00
(21.71)
58.21
(22.51)
48.90
(29.32)
Estimated consumption (%)
1.37
2.71
3.96
22.90
9.63
7.47
2.30
36.92
Panel B: Building characteristics and heating methods
Heated area (\({\mathbf{m}}^{2})\)
1665.10
(2231.07)
2 347.93
(2551.56)
2 444.07
(2646.94)
2 301.77
(2532.87)
630.54
(1162.76)
941.47
(1660.61)
1 215.67
(1799.33)
955.40
(1389.56)
Average apartment size
(\({\mathbf{m}}^{2})\)
90.04
(36.95)
93.76
(26.66)
92.58
(97.89)
82.26
(23.26)
83.96
(29.43)
85.73
(25.95)
89.73
(31.81)
80.75
(26.35)
District heating (%)
94.62
97.68
99.19
95.92
Combustion (%)
5.38
2.32
0.81
3.08
Heat pumps (%)
84.54
91.10
89.40
90.12
Other electric heating (%)
15.47
8.90
10.60
9.88
Panel C: Municipal variables
District heating price (SEK/sqm)
162.18
(14.31)
161.24
(14.07)
162.57
(14.07)
163.91
(14.90)
Electricity price (SEK/sqm)
94.55
(13.84)
88.16
(13.02)
89.37
(12.36)
98.29
(12.09)
Income (tSEK)
309.37
(41.27)
295.72
(42.54)
302.72
(37.57)
311.22
(35.63)
317.95
(47.75)
300.74
(47.69)
308.80
(46.30)
314.97
(37.83)
Number of buildings (observations)
2 696
1 549
985
2 306
737
281
566
891
In Panel A, estimated consumption (%) represents the share of EPCs based on estimated energy consumption values, rather than measured values. In Panel C, a small share of observations have missing values for energy prices. The average district heating/electricity price across the same region and for the same EPC approval year is used to impute the missing values
Panel A summarizes the specific energy use in each sample group. Regardless of the type of heating method used, the mean values suggest a decreasing trend in specific energy use across the construction period groups. The standard deviation values indicate that groups with a larger number of buildings show greater variance in specific energy use compared to groups with fewer buildings. Furthermore, the statistics reveal that a relatively small proportion of specific energy use data is based on estimated values for buildings in the baseline group, Reform 1 group, and Reform 2 group. However, for recently constructed buildings in the Reform 3 group, the proportion of estimated energy use—rather than being based on actual measurements (energy bills)— has undergone a dramatic increase. Specifically, 22.90% of specific energy use for mainly district heating and 36.92% for electric heating are based on estimated values.
Panel B presents summary statistics on building characteristics and heating methods. The mean values on heated areas indicate that, for both types of buildings, relative to those in the baseline group, buildings constructed after the implementation of the performance-based regulation tend to be larger on average. The standard deviation values for heated areas suggest greater variation. The average apartment size for newly constructed buildings in the Reform 3 group is smaller than that of buildings constructed earlier. Furthermore, the majority of multi-apartment buildings relies on district heating, and most electrically heated buildings utilize heat pumps. Panel C shows that energy prices and income at the municipal level do not vary substantially across the four construction period groups.
Models and estimation methods
Ordinary Least Squares (OLS) and quantile regressions
To assess the effect of the introduced and tightened performance-based building codes on actual energy use, I compare energy use in multi-apartment buildings constructed under different building code standards. The regression model is estimated as specified below:
where \({energy\_use}_{i}\) is the specific energy use for building \(i\); \({R1}_{i}\) is the indicator variable for the introduced performance-based building codes in July 2006, which equals 1 for buildings in the Reform 1 group; \({R2}_{i}\) is the indicator variable for the first tightened building code, which equals 1 for buildings in the Reform 2 group; \({R3}_{i}\) is the indicator variable for the second tightened building code, which equals 1 for buildings in the Reform 3 group; \({X}_{i}\) represents a vector of control variables including heated area, average size of apartments, and energy prices; \(t\) is the time trend over year-of-construction; \({\alpha }_{m}\) is a vector of municipal dummies; \({estimated}_{i}\) is a dummy variable, which equals 1 for building \(i\) is based on estimated energy consumption values; and \({\varepsilon }_{i}\) is the error term.
In the regression, only multi-apartment buildings using district heating are included in the sample for buildings with a heating method other than electric heating.10 The model is estimated separately for buildings with district heating and for those with electric heating. Moreover, each model is estimated twice: Model 1 is based on observations with measured energy consumption, and Model 2 is based on observations that include both measured and estimated energy consumption, with \({estimated}_{i}\) included in the model.
OLS estimation with robust standard errors is conducted. The primary parameters of interest are \({\beta }_{1}\), \({\beta }_{2}\), and \({\beta }_{3}\). \({\beta }_{1}\) captures the effect of the introduced performance-based building codes by comparing specific energy use for buildings in the Reform 1 group relative to specific energy use for buildings in the baseline group. Moreover, \({\beta }_{2}\) and \({\beta }_{3}\) capture the effects of the tightened performance-based building codes by comparing energy use for buildings in the Reform 2 group and the Reform 3 group respectively, relative to energy use for buildings in the baseline group. If the tightened building codes in later periods have an actual effect on reducing energy use, we should expect the absolute value of \({\beta }_{3}\) to be larger than \({\beta }_{2}\), and the absolute value of \({\beta }_{2}\) to be larger than \({\beta }_{1}\).
Furthermore, the effects of changes in building codes may vary depending on the levels of specific energy use. Specifically, for buildings that actually consume less energy than the required level, indicating buildings with better energy performance, the effect of the regulation might be relatively small.11 Conversely, for buildings where the actual energy use exceeds the required level, indicating buildings with worse energy performance, the impact of the building codes could be more substantial.
Considering the feature of voluntary higher efficiencies in buildings, the quantile regression can be applied to identify heterogeneous effects across different levels of energy use. Unlike the OLS estimation, quantile regression encounters convergence issues due to the large number of municipal dummies. To address this, regional dummies are used instead of municipal dummies to control for other unobserved factors.
Structural break analysis
Given the empirical challenge in determining the lag between the issuance of building permits and the completion of construction, there is an uncertainty of either overestimating or underestimating the effects of building codes. A complementary strategy to identify the impact of building codes is to assess whether there exists a shift in the trend towards greater energy efficiency following the introduction of performance-based building codes. To determine if the slope of the time trend across years of construction becomes steeper with the adoption of performance-based building codes, a structural break model is estimated as below:
where \({R}_{i}\) represents the indicator variable for performance-based building codes, which equals 1 for buildings in all Reform groups with performance-based building codes and 0 for buildings in the baseline group. \(({R}_{i}*t)\) is the interaction term between the reform indicator and the time trend.
The primary parameter of interest in this analysis is \(\sigma\), which provides a direct comparison of the time trend when (\({R}_{i}=1\)) relative to when (\({R}_{i}=0\)). Moreover, the Chow test is used to test for the presence of a structural break following the introduction of performance-based building codes. Specifically, it examines whether the time trend coefficients after the implementation of performance-based building codes (\({R}_{i}=1\)) differ from those before the implementation (\({R}_{i}=0\)).
Nearest-neighbor matching on causal effects
In addition to the overall reform effects, a specific aspect of the tightened regulation is that the required change levels vary across Sweden. For instance, in the stricter building codes (BFS, 2015:3 – BBR22), the required tightened energy efficiency level was set 5 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) higher in the new climate zone IV than climate III for buildings with both district heating and electric heating.12 If the stricter building codes in 2015 (Reform 3) have a causal influence on energy use, the change in energy use before and after Reform 3 should be larger for multi-apartment buildings at the border of climate zone IV than for those at the border of climate zone III. To test this hypothesis, the model is estimated as follows:
where \({\left\{reform\_border\right\}}_{i}\) is the interaction term between Reform 3 and climate zone border indicators. Specifically, the reform indicator equals to 1 for buildings in the Reform 3 group (after the reform) and 0 for buildings in the Reform 2 group (before the reform); the border indcator equals to 1 for buildings at the border of climate IV (treatment group) and equals to 0 for buildings at the border of climate zone III (control group).13\({X}_{i}\) refers to a vector of time-invariant covariates for matching. In order to match the buildings before and after Reform 3 that may be managed by the same housing company or share similar social and environmental conditions, the postal code, instead of municipal dummy variables, is used for matching. Furthermore, time-invariant building characteristics are also selected as covariates.
The nearest-neighbor matching estimates the average treatment effect on the treated (ATT) based on this design. \(\beta\) in Eq. (3) is the primary interested parameter, which compares the change in energy use for buildings at the border of climate zone IV after Reform 3 to before, relative to the change in energy use for buildings at the border of climate zone III. Nearest-neighbor matching eliminates the need for functional-form assumptions, making the estimator more flexible. The method determines the nearest by using a weighted function based on the covariates for each observation.14 Moreover, given that building characteristics and postal codes are selected as continuous variables, the bias-corrected estimator is employed to account for large-sample bias (Abadie & Imbens, 2011).
Results
Overall compliance with BBR requirements
Before reporting results from the regression analysis, an overall statistical result on whether the actual energy use meets the BBR requirements are presented. Figure 3 provides an overview of the ratio between specific energy use and the required value in BBR, indicating compliance with BBR regulations for each construction group. A ratio lower than one indicates that the actual energy use is below the required level. Although there is no requirement for buildings in the baseline group, a ratio is also calculated based on the requirements for buildings in the Reform 1 group to enable a comparison of regulation compliance before and after the introduction of performance-based building codes in July 2006.
Fig. 3
The ratio between specific energy use and BBR requirements by group of construction years (sample only having measured energy consumption) (a) Buildings that mainly have district heating. Notes: The cumulative percentages of the ratio that is below 1 in these four groups are 33.9%, 76.4%, 63.7%, and 53.4%, respectively. (b) Buildings that have electric heating. Notes: The cumulative percentages of the ratio that is below 1 in these four groups are 83.2%, 95.8%, 57.7%, and 69.0%, respectively. In both panels, for multi-apartment buildings in the baseline group, the ratio is calculated as actual specific energy use divided by the requirements for buildings in Group 1
For multi-apartment buildings that mainly have district heating, Fig. 3a indicates that the mean ratio for buildings in the baseline group is 1.13, which is higher than the implemented regulated level. In the case of buildings constructed in the Reform 1 group, the overall actual energy use falls below the requirement value, with an average ratio of 0.86. Moreover, following the introduction of performance-based building codes, the cumulative percentage of actual energy use meeting the regulatory requirements increases significantly, rising from 33.9% to 76.4%. For buildings constructed after the regulation tightening, the mean ratios for the Reform 2 group and the Reform 3 group are 0.94 and 1.04, respectively. This suggests that, on average, compliance meets the regulated level across newly constructed buildings subject to the building codes. However, the distribution reveals that only 63.7% and 53.4% of sampled buildings in these two groups have energy use below the regulatory level, indicating room for ongoing improvement in energy efficiency.
Concerning buildings with electric heating, Fig. 3b shows that the actual energy use for buildings in the baseline group is lower than the implemented regulated level in 2006, with an average ratio of 0.73. This is because there are no stricter requirements for multi-apartment buildings with electric heating, making it easier to achieve the same requirement than for non-electric buildings. Despite the absence of specific stricter requirements, the reduction in the average ratio (from 0.73 to 0.62) and the increase in the proportion of actual energy use below regulatory requirements (from 83.2% to 95.8%) suggest that the performance-based regulation may have an effect in saving energy for buildings with electric heating. For newly constructed electrical heated buildings in the Reform 2 group and the Reform 3 group, the results show a pattern similar to that for buildings mainly with district heating. While the overall actual energy use nearly meets the requirement levels, with average ratios per group of 1.00 and 1.02 per group, the ratio distributions indicate potential for energy efficiency improvement, as only 57.7% and 69.0% of newly constructed electrically heated buildings in these two groups meet the regulatory requirements.
Results from OLS estimation
Table 4 presents the OLS-estimated coefficients for Model (1). For buildings with district heating, the results show that the introduction of performance building codes (Reform 1) is associated with a reduction of 17.89 \(\text{kWh}/{\text{m}}^{2}\) per year in specific energy use. Regarding the effect of the tightened performance-based building codes, the results suggest that relative to buildings in the baseline group, Reform 2 and Reform 3 cause reductions in specific energy use of approximately 20.46 \(\text{kWh}/{\text{m}}^{2}\) per year and 18.23 \(\text{kWh}/{\text{m}}^{2}\) per year, respectively.
Table 4
Effects of building regulation reforms on specific energy use (OLS estimation)
Dependent variable:
Specific energy use
(kWh/m2 per year)
Buildings with district heating
Buildings with electric heating
Model (1)
Model (2)
Model (1)
Model (2)
Reform 1
−17.89***
−18.872***
−7.84***
−5.546**
(2.00)
(1.952)
(2.831)
(2.615)
Reform 2
−20.462***
−22.271***
−13.843***
−10.382**
(2.937)
(2.85)
(4.774)
(4.453)
Reform 3
−18.228***
−21.386***
−10.861
−7.622
(3.737)
(3.631)
(6.638)
(6.153)
Heated area (m2)
−0.002***
(0.0001)
−0.002***
(0.0002)
−0.001**
(0.001)
−0.001***
(0.001)
Average apart. size (m2)
−0.067**
(0.029)
−0.066**
(0.028)
−0.216***
(0.033)
−0.195***
(0.028)
District heating prices
0.056
0.1
(0.073)
(0.066)
Electricity prices
−0.068
−0.066
(0.072)
(0.062)
Time trend
−1.619***
(0.262)
−1.426***
(0.253)
−0.724
(0.441)
−1.023**
(0.411)
Estimated dummy
−18.147***
(0.835)
−10.606***
(1.375)
Municipality dummies
Yes
Yes
Yes
Yes
Observations
6847
7493
2019
2452
R-squared
0.476
0.526
0.513
0.544
*** and ** denote significance at the 1% and 5% significance level, respectively. Robust standard errors are in parentheses. Estimated coefficients of municipal dummies are not reported. Reform 1 is the indicator variable for the introduction of performance-based building codes in 2006; Reform 2 is the indicator variable the first tightened building codes (in 2009 for buildings with electric heating and in 2012 for those with district heating); Reform 3 is the indicator variable the second tightened building codes in 2015 for both types of buildings. Wald tests are used to check whether the difference between the coefficients on Reform 2 and Reform 1 are statistically significant. The p-value of the test for buildings with district heating is 0.0608, and the p-value for buildings with electric heating is 0.046
Additionally, the descriptive statistics show that the average specific energy use for the baseline group is 126.34 \(\text{kWh}/{\text{m}}^{2}\) per year. Therefore, the estimated absolute reduction amounts indicate that, for district-heated buildings, Reform 1, Reform 2, and Reform 3 lead to reductions of approximately 14.2%, 16.2%, and 14.4% in specific energy use, respectively. The results indicate that the first tightened building codes (Reform 2) led to an approximately 2% improvement in energy efficiency. Wald tests indicate that the increase is statistically significant at the 10% level. However, there is no evidence that the second tightened building codes have an actual effect on reducing energy use.
The finding could be attributed to the time trend across years of construction considered in the estimation. The results show a decreasing trend in energy use of about 1.62 \(\text{kWh}/{\text{m}}^{2}\) per year across construction years, corresponding to a reduction of about 1.57% per year.15 Moreover, when both measured and estimated energy consumption are included, the results for Model (2) show that, as the magnitude of the coefficient on time trend decreases, the effects of all reforms are slightly larger compared to results from Model (1).
For electrically heated multi-apartment buildings, the results for Model (1) suggest that Reform 1 leads to an approximate reduction of 7.84 \(\text{kWh}/{\text{m}}^{2}\) per year in specific energy use. Moreover, the results indicate that Reform 2 causes a reduction of 13.84 \(\text{kWh}/{\text{m}}^{2}\) per year in specific energy use, and that Reform 3 is associated with a reduction of 10.86 \(\text{kWh}/{\text{m}}^{2}\) per year. The effect of Reform 3 is statistically insignificant.
Considering the average specific energy use for buildings in the baseline group (80.93 \(\text{kWh}/{\text{m}}^{2}\) per year), the estimations indicate that, in comparison to buildings in the baseline group, buildings in the Reform 1, Reform 2, and Reform 3 groups undergo respective reductions of approximately 9.7%, 17.1%, and 13.4% reduction in specific energy use. The findings indicate that the first tightened building codes (Reform 2) had a more substantial impact, leading to an approximately 7.4% improvement in energy efficiency. Wald tests indicate that the increase is statistically significant at the 5% level. Nevertheless, the results do not show that Reform 3 (the further tightening of building codes) generates a larger effect on energy savings than does Reform 2.
The results show that the effect of the time trend on specific energy use for electrically heated buildings is statistically insignificant in Model (1). However, the negative value of 0.72 \(\text{kWh}/{\text{m}}^{2}\) per year suggests a trend towards increased energy efficiency, corresponding to an improvement of approximately 1.09% per year.16 Moreover, when estimated energy consumption is included in the estimation for Model (2), the effect of the time trend increases in size and becomes statistically significant at the 5% significance level, which results in the magnitude of the reform effects becoming smaller.
In addition to the reform and time trend variables, the results in Model (2) indicate that, when estimations are based on the whole sample, the dummy variable for estimated energy consumption is a significant factor. The result suggests that, relative to specific energy use data based on measured values, energy use based on estimations is approximately 18.15 \(\text{kWh}/{\text{m}}^{2}\) per year lower for buildings with district heating. Similarly, in comparison with measured energy use, the estimated value is 10.61 \(\text{kWh}/{\text{m}}^{2}\) per year lower for electrically heated buildings.
For other control variables, the results suggest that specific energy use is negatively associated with the size of the heated area and the average apartment size in the buildings. Energy prices do not have a significant negative effect on specific energy use, possibly because the energy use data in the EPCs are adjusted to normalized values, which are less sensitive to fluctuations in energy prices. Moreover, the positive coefficient of district heating prices is under expectation. Most district heating companies are municipally owned, which gives them the potential to be natural monopolies.
In both Models (1) and (2), municipality dummies are included to control for unobserved factors that may influence energy use at the municipal level. As a robustness check, a multilevel mixed-effects regression is also estimated to examine the effects of the reform on specific energy use, while accounting for variation at the municipality level (Raudenbush & Bryk, 2002). The results are reported in Appendix 2. The estimated coefficients from the multilevel regression generally align with those presented in Table 4. Additionally, the variance for municipality ID suggests that there is significant variability in specific energy use across municipalities.
Results from quantile regression
Figure 4 depicts the results of reform effects from quantile regressions with 95% confidence intervals. Specifically, for buildings with district heating, Fig. 4a shows that the effects of Reform 1 on reducing energy use are around 12–13 \(\text{kWh}/{\text{m}}^{2}\) per year between the 10 th and 40 th percentiles and approximately 15–20 \(\text{kWh}/{\text{m}}^{2}\) per year between the 50 th and 80 th percentiles, increasing to 27 \(\text{kWh}/{\text{m}}^{2}\) per year at the 90 th percentile for specific energy use. Furthermore, the magnitude of effects of both tightened building codes (Reforms 2 and 3) tends to increase with the quantile level, ranging from approximately 5 \(\text{kWh}/{\text{m}}^{2}\) per year at the 10 th percentile to 35 \(\text{kWh}/{\text{m}}^{2}\) per year at the 90 th percentile for specific energy use. When comparing the magnitudes of effects between the initial performance-based building code (Reform 1) and the tightened building codes (Reforms 2 and 3), it is observed that Reforms 2 and 3 have a larger impact than Reform 1 starting from above the 50 th percentile, with the difference becoming more pronounced at higher quantile levels.
Fig. 4
Estimated coefficients of quantile regressions with 95% confidence intervals (a) Buildings with district heating (b) Buildings with electric heating. Notes: Regressions based on specific energy use without estimated consumptions. Regional dummies instead of municipal dummies are included in the estimation. The horizontal dash lines represent OLS estimates with 95% confidence intervals
For electrically heated buildings, Fig. 4b indicates that Reform 1 does not lead to a reduction in energy use in the lower and middle quantile ranges. However, a significant impact of Reform 1 is observed for the most energy-inefficient electrically heated buildings, with the peak effect being around 24 \(\text{kWh}/{\text{m}}^{2}\) per year at the 90 th percentile. This finding is in line with expectations, considering that the introduction of the performance-based building codes (Reform 1) did not impose stricter requirements for electrically heated apartment buildings as compared to those with district heating. Consequently, Reform 1 is primarily effective in reducing energy use in poorer-performing buildings. Similarly, the impacts of the tightened building codes (Reforms 2 and 3) are moderate below the 80 th percentile but become more significant at higher quantile levels for specific energy use, with the peak effects being approximately 33 \(\text{kWh}/{\text{m}}^{2}\) per year for Reform 2 and 30 \(\text{kWh}/{\text{m}}^{2}\) per year for Reform 3.
In summary, for both buildings with district heating and electric heating, it is evident that the peak of energy-saving effects from both the introduced and the tightened performance-based building codes occurs at approximately the 90 th percentile. Furthermore, the predicted energy use trends across construction years for different quantiles are presented in Appendix 3. The statistics suggests that the mean predicted energy use at the 90 th percentile is around 135 \(\text{kWh}/{\text{m}}^{2}\) per year for district-heated buildings and 97 \(\text{kWh}/{\text{m}}^{2}\) per year for electrically heated buildings, exceeding the highest regulated requirements (130 \(\text{kWh}/{\text{m}}^{2}\) per year and 95 \(\text{kWh}/{\text{m}}^{2}\) per year). Therefore, the findings from the quantile regressions indicate that the impact of building codes is more substantial for buildings where the actual energy use exceeds the mandated levels.
Results from structural break analysis
Table 5 presents the coefficients of the interaction term between the time trend and the reform indicator. For buildings with district heating, the results suggest that there is an additional decrease in specific energy use by 0.92 \(\text{kWh}/{\text{m}}^{2}\) per year after the implementation of performance-based building codes. In the case of electrically heated buildings, the difference in the slope of the time trend is even more pronounced. After the implementation, specific energy use decreases by an additional 1.45 \(\text{kWh}/{\text{m}}^{2}\) per year. Furthermore, the Chow test rejects the null hypothesis of no change in the slope of time trend between the two groups. The results indicate that there is a shift in the trend towards significantly greater energy efficiency after the implementation of regulation.
Table 5
Coefficients on the interaction term between the time trend and the reform indicator
Dependent variable:
Specific energy use
(kWh/m2 per year)
Buildings with district heating
Buildings with electric heating
Trend * Reform
−0.923***
(0.343)
−1.450**
(0.604)
Chow test (F-test)
7.23
[0.007]
5.77
[0.016]
In the first panel, *** and ** denote the significance at 1% and 5% levels, respectively. Robust standard errors are in parentheses. In the second panel (for the Chow test), the p-value is indicated in brackets. The p-value is less than the 1% for buildings with district heating and less than 5% for buildings with electric heating
To further visualize the change in the time trend, a robustness regression of specific energy use over each construction year, with controls for other factors, is estimated. In the case of buildings with district heating, Fig. 5a shows a steeper downward trend in specific energy use after the introduction of performance-based building codes. However, for buildings with electric heating, Fig. 5b shows a slightly positive trend in specific energy use prior to the reform, leading to a large difference in the slope of the time trend.17 The findings confirm the shift in the slope of the time trend before and after the implementation of performance-based regulation.
Fig. 5
Differences in mean specific energy use by year of construction. (a) Buildings with district heating (b) Buildings with electric heating. Notes: Results are based on the model: \({energy\_use}_{i} = {\beta }_{0}+{\theta year\_of\_construct}_{i}+{\gamma X}_{i}+{\alpha }_{m}+{\varepsilon }_{i}\). The estimation controls for \({X}_{i}\) (observed building characteristics and energy prices) and \({\alpha }_{m}\) (unobserved municipal-specific factors). Coefficients \(\theta\) are plotted. For buildings with district heating, the construction year 2008 is the omitted base time (as the introduced performance-based building codes in July 2006 may have an effective impact for buildings constructed since 2009), and for buildings with electric heating, the construction year 2007 is the omitted base time (as the introduced performance-based building codes in July 2006 may have an effective impact for buildings constructed since 2008)
Figure 6 indicates the effects of Reform 3 on specific energy use for buildings at the border of climate zone IV with higher tightened requirements compared to those at the border of climate zone III with relatively lower requirements. For district-heated buildings, Reform 3 is associated with a reduction in specific energy use by approximately 4.21 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) more for buildings at the border of climate zone IV, relative to those at the border of climate zone III. For electrically heated buildings, Reform 3 leads to a decrease in specific energy use by approximately 8.48 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) more for buildings at the border of climate zone IV. The additional effects are statistically significant for both types of buildings. Compared to the targeted gap of 5 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\), the robustness checks provide evidence that the stricter building codes have a causal effect on energy use for buildings with district heating and for those with electric heating.
Fig. 6
Comparison of Reform effect across climate zones. Notes: Robust Abadie-Imbens standard errors are used in the estimation. For each estimation, the number of buildings in the treatment group, the number of matched control buildings for each treated one, and the diagnostics of covariate balance between treated and matched control observations are reported in Appendix 4
The effects of building regulation on energy efficiency
This study differs from previous empirical studies in several respects. Previous studies have primarily concentrated on one particular program, such as either the introduction of a given building code (Aroonruengsawat et al., 2012; Holian, 2020; Novan et al., 2022) or a change in the code (Jacobsen & Kotchen, 2013; Kellogg & Cumbre-Gibbs, 2023; Kotchen, 2017). Based on the characteristics of building regulations in Sweden, this study examines the impact of both introduced and stricter building performance-based building codes on energy savings for district heating and electric heating. Furthermore, while most previous studies rely on survey data at the housing level, this study utilizes measured energy use data from EPCs, allowing for the inclusion of most buildings in the building stock within the study sample.
The results from the OLS estimation suggest a notable trend toward greater energy efficiency, estimated at 1.57% \(\text{per year}\) for buildings with district heating and 1.09% \(\text{per year}\) for electrically heated buildings. The finding supports the statement that the absolute energy use in buildings has fallen by nearly 1% per year in Sweden (Tammiste et al., 2018).
After accounting for this trend, the results indicate that the introduction of performance-based building codes in July 2006 (Reform 1) resulted in a 14.2% increase in energy efficiency for buildings with district heating and a 9.7% increase for buildings with electric heating. Moreover, the first tightened building codes (Reform 2) generates an additional 2% increase in energy efficiency for district-heated buildings and an approximately 7.4% improvement in energy efficiency for electrically heated buildings. However, there is no evidence suggests the second tightened building codes (Reform 3) further enhanced energy efficiency. The findings on Reforms 1 and 2 are consistent with Hjortling et al. (2017), whose study suggest that, along with the stricter building codes, the measured energy consumption has been lowered compared to multi-apartment buildings built earlier in Sweden.
Compared to cases in the US, the finding that the implementation of performance-based building codes positively impacts energy efficiency is consistent with several empirical studies (e.g., Aroonruengsawat et al., 2012; Levinson, 2016; Novan et al., 2022). However, there is heterogeneity in the types of energy and buildings assessed in these various studies. Nevertheless, the effects of introducing building codes on electric heating observed in this study are smaller compared to the effects observed in California houses for electricity savings (Levinson, 2016), but are similar to those from Novan et al. (2022). However, the impact of the first tightening of the building code (Reform 2) on electric heating is larger than the electricity savings observed due to the change enacted in building codes in Florida (Jacobsen & Kotchen, 2013; Kotchen, 2017).
Unlike previous empirical studies, quantile regression is also applied in this study, considering the feature of voluntary higher efficiencies in buildings. The results indicate that the regulatory effects are relatively modest for buildings with better energy performance and more substantial for buildings where the actual energy use exceeds the required standards. Furthermore, the structural break analysis in this study indicates that the slope of the time trend becomes steeper following regulation implementation. A plausible explanation for the shift might be that adoption of performance-based building codes has increased the adoption of energy efficiency measures.
The actual energy savings in comparison to BBR requirements
In an overview of the ratios between specific energy use and the required values in BBR, the statistics suggest that compliance nearly meets the regulated levels across constructed buildings subject to the building codes, with the average ratio below or close to 1. However, among newly constructed buildings following the implementation of the stricter building codes (Reform 2 and Reform 3), the distribution of the ratios indicates that just over half of the sampled buildings have levels of energy use below the regulated level. In addition, there appears to be a trend towards lower compliance. This suggests that there is room for improvements in compliance with energy efficiency requirements.
In addition, the actual energy savings magnitude from the matching method can be compared with the saving magnitudes stipulated by tightened BBR requirements. For example, the required tightened levels for buildings with district heating were 20 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in 2012 (BFS, 2011:26 – BBR19) and 10–15 \(\text{kWh}/{\text{m}}^{2}\text{ per year}\) in 2015(BFS, 2015:3 – BBR22). However, the actual saving effects are significantly smaller relative to the BBR requirements.
Hjortling et al. (2017) suggest that newly constructed multi-dwelling buildings and other commercial buildings often have higher energy use levels than required by the building codes. The results in this study, suggesting that actual energy savings are smaller than those stipulated by tightened requirements in regulations, are in line with the arguments by Hjortling et al. (2017). The finding is also consistent with those in Levinson (2016), who observed that the achieved energy savings were substantially less than the projected ones in regulatory context.
Other implications
In addition to the methodological implications regarding the importance of accounting for the time trend based on years of construction when evaluating the effects of continued increases of stringency in the building codes, the findings in this study also generate two additional policy implications.
Approximately more than half of the new buildings in the sample were constructed with higher energy efficiency than the regulatory requirements. Although the minimum energy requirements are not binding, one could expect some efficiency gains due to the incentives from the energy performance classes. However, the results from the quantile regression indicate that the regulatory effects on energy efficiency improvement are relatively modest for these buildings. This indicates that, in addition to regulatory instruments, other policy instruments could be considered to monitor and encourage an increase in actual energy efficiency in these high-performing buildings.
Furthermore, most empirical analyses in this study rely on measured energy use values. However, a small portion of the energy use data from EPCs is based on estimated values, particularly for newly constructed buildings where energy usage cannot be metered. In Sect. "Results from OLS estimation" (Model 2), the study examines the difference in energy use between constructions with estimated and measured energy use values. The findings indicate that, compared to specific energy use based on measured values, the estimated energy use is, on average, 18.15 \(\text{kWh}/{\text{m}}^{2}\) per year lower for district-heated buildings and 10.61 \(\text{kWh}/{\text{m}}^{2}\) per year lower for electrically heated buildings.
This finding provides evidence of the energy performance gap, with the observed difference in energy use values being both economically and statistically significant. The result is in line with those of previous empirical studies (e.g., Filippidou et al., 2019; Hörner & Lichtmeß, 2019). It suggests that engineers might be over optimistic about the gains in energy efficiency. One implication of this finding is that engineering experts may need to investigate why estimated energy use differs so much from actual energy use and whether the gap tends to be larger for energy-efficient buildings (labelled A or B) (Filippidou et al., 2019; Majcen et al., 2013).
Limitations and future research
One significant empirical challenge in the study arises from the lag between building permits and constructions. The sample design only relied only on the expected average construction time. Despite excluding buildings constructed in the transition period, there remains the possibility of overestimating or underestimating the effects of regulation. An ideal approach for future studies would be to combine the date information from building permits with current EPC data. In addition, the study uses specific energy use as the outcome variable in order to ensure a fair comparison. However, changes in the measurement of energy performance on EPCs are not captured in the results. Another limitation of the study is the absence of household characteristics, such as average family size and household income, in the data used. Future research could enhance the depth of analysis by incorporating information about building owners and tenants.
Furthermore, this study focuses exclusively on residential apartment buildings, with future research opportunities in examining energy use in other types of buildings, such as small houses or commercial buildings. Given the policy of voluntary higher efficiencies and the wide variation in the ratios between specific energy use and the required values, a potential avenue for further study is identifying the determinants of regulation compliance. Additionally, for a more comprehensive understanding of the energy performance gap, future studies could explore methods for quantifying the gap and investigating its determinants within the Swedish building sector.
Acknowledgements
The author acknowledges financial support provided by the Foundation for Baltic and East European Studies (Östersjöstiftelsen).
Declarations
Ethical approval and consent to participate
Not applicable, as the research does not involve human or personal data.
Human ethics
Not applicable, as the research does not involve human subjects.
Consent for publication
Not applicable, as the research does not include personal data.
Competing interest
The author declares no competing interests or personal relationships that could have appeared to influence the work.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Estimation of Model (1) using Multilevel mixed-effects regression
Dependent variable:
Specific energy use
(kWh/m2 per year)
Buildings with district heating
Buildings with electric heating
Model (1)
Model (2)
Model (1)
Model (2)
Reform 1
-17.784***
-18.811***
-7.47***
-5.564**
(1.731)
(1.656)
(2.819)
(2.534)
Reform 2
-20.546***
-22.359***
-13.542***
-10.802***
(2.58)
(2.458)
(4.321)
(3.894)
Reform 3
-18.103**
-21.29***
-11.591*
-9.03*
(3.215)
(3.069)
(6.085)
(5.452)
Heated area (m2)
-.002***
(0.0001)
-.002***
(0.0001)
-.001**
(.0004)
-.001***
(.0003)
Average apart. size (m2)
-.068**
(.006)
-.068**
(.006)
-.227***
(.021)
-.173***
(.017)
District heating prices
.002
.034
(.048)
(.045)
Electricity prices
-.067
-.061
(.059)
(.05)
Time trend
-1.618***
(.222)
-1.424***
(.211)
-.763*
(.408)
-.994**
(.366)
Estimated dummy
-18.704***
(1.091)
-11.226***
(1.51)
Municipality ID
var(constant)
185.31***
(26.16)
174.93***
(23.99)
310.98***
(48.76)
302.70***
(44.21)
Observations
6847
7493
2019
2452
***, **, and * denote significance at the 1%, 5%, and 10% significance level, respectively. Reform 1 is the indicator variable for the introduction of performance-based building codes in 2006; Reform 2 is the indicator variable the first tightened building codes (in 2009 for buildings with electric heating and in 2012 for those with district heating); Reform 3 is the indicator variable the second tightened building codes in 2015 for both types of buildings
The other two targets are related to renewable energy and emission reduction. Broadly, energy efficiency is often defined by the simple ratio between useful output of a process and energy input into a process (Patterson, 1996).
Construction quality refers here to, for example, the use of energy efficiency systems and the use of quality construction materials, while physical attributes refer to e.g. housing type and size.
Due to newly introduced parameters in the energy declaration form, energy performance is measured with three methods: Up to and including 2018-12-31, energy performance is measured as specific energy use; between 2019-01-01 and 2020-08-31, energy performance is measured as the primary energy number with primary energy factors (updated in BBR25); and from 2020-09-01, it is measured as the primary energy number with weighting factors (updated in BBR29).
Buildings constructed in both 2007 and 2008 were excluded due to the fact that the relevant changes came into force in the second half of 2006 and with a transition period until July 2007.
For example, for multi-apartment buildings mainly with district heating constructed in 2000–2019, the average heating areas is 2102 \({\text{m}}^{2}\), with 24 apartments in the groups. For electrically heated buildings, the average heating area is 901 \({\text{m}}^{2}\) and the number of apartments is 11.
The presence of different levels of energy classes (labeled A-G) can affect the incentives to increase energy performance above the minimum requirements.
For buildings with district heating, the energy use requirements were tightened by 10 \(\text{kWh}/{\text{m}}^{2}\) per year in climate zone III and by 15 \(\text{kWh}/{\text{m}}^{2}\) per year in climates zone IV. For buildings with electric heating, requirements were tightened by 5 \(\text{kWh}/{\text{m}}^{2}\) per year in climate zone III and by 10 \(\text{kWh}/{\text{m}}^{2}\) per year in climate IV.
The border regions of climate zone IV include 08 Kalmar, 10 Blekinge, 12 Skåne, and 13 Halland; the border regions of climate zone III include 05 Östergötland, 06 Jönköping, 07 Kronoberg, and 14 Västra Götaland.
The positive trend prior to the reform tends to disappear if the starting time is 1990.
Abadie, A., & Imbens, G. W. (2011). Bias-corrected matching estimators for average treatment effects. Journal of Business and Economic Statistics,29(1), 1–11.MathSciNetCrossRef
Aroonruengsawat, A., Auffhammer, M., & Sanstad, A. H. (2012). The impact of state level building codes on residential electricity consumption. Energy Journal,33(1), 31–52.CrossRef
Davis, L. W., Martínez, S., & Taboada, B. (2018). How Effective is Energy-efficient Housing? Evidence From a Field Experiment in Mexico. IDB working paper, IDB-WP-843.
Filippidou, F., Nieboer, N., & Visscher, H. (2019). Effectiveness of energy renovations: A reassessment based on actual consumption savings. Energy Efficiency,12(1), 19–35.CrossRef
Hjortling, C., Björk, F., Berg, M., & Klintberg, T. (2017). Energy mapping of existing building stock in Sweden: Analysis of data from Energy Performance Certificates. Energy and Buildings,153(341), 355.
Holian, M. J. (2020). The impact of building energy codes on household electricity expenditures. Economics Letters,186, 108841.CrossRef
Hörner, M., & Lichtmeß, M. (2019). Energy performance of buildings: A statistical approach to marry calculated demand and measured consumption. Energy Efficiency,12(1), 139–155.CrossRef
Jacobsen, G. D., & Kotchen, M. J. (2013). Are building codes effective at saving energy? Evidence from Residential Billing Data in Florida, the Review of Economics and Statistics,95(1), 34–49.
Kellogg, K. A., & la Cumbre-Gibbs, N. (2023). The impact of state level residential building code stringency on energy consumption in the United States. Energy and Buildings,278, 112607.CrossRef
Kotchen, M. J. (2017). Longer-run evidence on whether building energy codes reduce residential energy consumption. Journal of the Association of Environmental and Resource Economists,4(1), 135–153.CrossRef
Levinson, A. (2016). How much energy do building energy codes save? Evidence from California Houses, American Economic Review,106(10), 2867–2894.
Majcen, D., Itard, L., & Visscher, H. (2013). Theoretical vs. Actual energy consumption of labelled dwellings in The Netherlands: Discrepancies and policy implications. Energy Policy,54, 125–136.CrossRef
Mavrogianni, A., Johnson, F., Ucci, M., Marmot, A., Wardle, J., Oreszczyn, T., & Summerfield, A. (2013). Historic variations in winter indoor domestic temperatures and potential implications for body weight gain. Indoor and Built Environment,22(2), 360–375.CrossRef
Novan, K., Smith, A., & Zhou, T. (2022). Residential Building Codes Do Save Energy: Evidence From Hourly Smart-Meter Data. Review of Economics and Statistics,104(3), 483–500.CrossRef
Pagliaro, F., Hugony, F., Zanghirella, F., Basili, R., Misceo, M., Colasuonno, L., & del Fatto, V. (2021). Assessing building energy performance and energy policy impact through the combined analysis of EPC data – The Italian case study of SIAPE. Energy Policy,159, 112609.CrossRef
Pasichnyi, O., Wallin, J., Levihn, F., Shahrokni, H., & Kordas, O. (2019). Energy performance certificates — New opportunities for data-enabled urban energy policy instruments? Energy Policy,127, 486–499.CrossRef
Patterson, M. G. (1996). What is energy efficiency? Concepts, Indicators and Methodological Issues. Energy Policy,24(5), 377–390.CrossRef
SW Raudenbush AS Bryk 2002 Hierarchical Linear Models: Applications and Data Analysis Methods (2nd ed.). Thousand Oaks, CA: Sage
Saussay, A., Saheb, Y., & Quirion, P. (2012). The impact of building energy codes on the energy efficiency of residential space heating in European countries: A stochastic frontier approach, International Energy Program Evaluation Conference, 12–14.
Tammiste, L., Poltimäe, H., Kuldna, P., Kallaste, T., Kirsimaa, K., & Kuusk, O. (2018). Nordic Green to Scale 2 - CO2 abatement potential, cost, enablers, barriers and recommendations: Estonia, Latvia, Lithuania, Poland and Ukraine. Stockholm Environment Institute Tallinn Centre.
von Platten, J., Holmberg, C., Mangold, M., Johansson, T., & Mjörnell, K. (2019). The renewing of Energy Performance Certificates—Reaching comparability between decade-apart energy records. Applied Energy,255, 113902.CrossRef
