Determinants to the adoption of energy-efficient retrofits and the role of policy measures

ABSTRACT

Buildings are responsible for around 45% of total energy consumption and one-third of CO₂ emissions annually in Switzerland. Policymakers have argued that an increased rate of energy-saving retrofits in existing buildings will play a critical role in meeting the energy and climate policy goals. This article examines the determinants for the households’ investment decisions to undertake energy-efficient retrofits and investigates the role of policy measures. We estimate random effects probit models using a rich data for 1663 owned single-family Swiss households for retrofits undertaken during 2010 to 2014. Results suggest that while the building vintage appears to be a relevant determinant; energy-related attitudes of decision makers and policy instruments are also likely to play an important role. In particular, we find a suggestive evidence of tax deduction policy in supporting households’ decisions to undertake energy-saving retrofits as well as the intensity of renovation. Direct monetary policies such as tax deductions should be focused particularly upon poorly insulated old buildings and those that rely on non-renewable energy sources for heating purposes.

KEYWORDS:

• Renovation
• tax deduction
• energy efficiency
• Switzerland
• residential sector

JEL CLASSIFICATION:

• C23
• D04
• H23
• H71
• Q48

I. Introduction

Buildings account for around 40% of total energy consumption and CO₂ emissions in European countries, primarily due to poor insulation in existing buildings as well as continued dependence on fossil fuels for heating purposes. In Switzerland too, the buildings stock are responsible for approximately 45% of the total end energy demand and one-third of the overall CO₂ emissions (Energieschweiz 2022). Even though technological solutions exist to achieve substantial improvements, the current rate of energy-efficient renovations for countries in the European Union (EU), and in Switzerland, is only about 1–1.5% per year (Jakob et al. 2013) and an important increase is necessary to meet the energy and climate policy goals.¹

For home owners, several factors determine whether or not they undertake new energy-efficiency related investments. These include crucial attributes such as age of the building and level of income. Behavioural and cognitive factors such as energy-related knowledge and energy-saving behaviour may also be important. Given that the upfront costs of retrofits are likely to be high, and the (private) financial benefits on these investments – in term of reduced energy costs and increased property value – maybe realized only after few years, policy instruments become attractive and already play an important role in EU as well as in the US.² In Switzerland too, the government promotes energy-saving renovations primarily with three types of policy instruments: subsidy programs, availability of tax deductions for energy-efficiency investments, and support for energy-related audit and consulting. Our main objective in this study is to analyse the determinants to the adoption of energy-efficient retrofits (EERs) by Swiss households, and we are particularly interested in examining the role of policy instruments as potential drivers. We put particular emphasis on the tax deduction policy because i) a tax policy is generally more salient and owners are asked every year about it, e.g. when filing the tax returns, ii) tax deductions are also easier to obtain than a subsidy for an energy saving retrofitting – in fact, for a retrofitting subsidy, the owner of a single-family house generally has to request a form, fill-it in before starting the retrofitting work, and send all documents to an administrative unit that will evaluate the request, and iii) there are ongoing political discussion on introducing more tax credits for energy-efficient retrofits.

Few studies have looked into the effectiveness of policies that provide incentives for residential energy-efficiency renovations and have found mixed results. Tax incentives for energy efficiency appear to have existed in the US since the 1970s, but some of the first empirical studies did not find a positive effect on energy-efficiency investments (Walsh 1989; Dubin and Henson 1988). Hassett and Metcalf (1995), on the other hand, highlighted the importance of controlling for unobserved heterogeneity and exploit panel data in order to measure the impact of tax-based incentive policies for making energy conservation investments by US households. They find a positive and significant impact of tax-related policy for such investments. Alberini, Bigano, and Boeri (2014), who examine free-riding behaviour on energy efficiency incentives among Italian homeowners, also argue that when it comes to household responsiveness to the incentive amount, ‘the evidence is mixed and inconclusive’.³

We examine the determinants of i) investment decisions to undertake an EER, and ii) the renovation intensity, in owned single-family dwellings using a panel data from 2010 to 2014 for a sample of 1663 Swiss households. The results provide suggestive evidence that the tax deduction policy has a significant positive effect on households’ decisions to undertake energy-saving retrofits. Our methodological framework relates to that of Kok, McGraw, and Quigley (2011) and Banfi, Filippini, and Ramseier (2011), but we use richer data that also allows us to account for the role of behavioural and cognitive factors, in addition to socio-demographics and dwelling attributes. The article contributes to the relatively scarce literature that examines empirical evidence on the impact of policy measures, such as tax deductions and subsidies, on energy-saving renovation decisions. Notably, this is also a rather new study that uses a large panel of owned, single-family households and their investment decisions in energy-efficient retrofits – one that is likely to have resemblance to similar settings across many countries within Europe.

II. Methodology and data

Empirical model

The primary outcome of interest in our analysis is a dichotomous variable that captures whether or not a household has undertaken an EER. This variable is equal to one if a household performed one or more EER, and is zero otherwise.⁴ For this main outcome, we estimate a probit model of the following form:

(1) $\begin{array}{rl}& \mathit{EE}{R}_i^{\ast }={\beta }_{\mathit{SEC}}\mathit{SE}{C}_i+{\beta }_{\mathit{DW}}\mathit{D}{W}_i+{\beta }_{\mathit{LIT}}\mathit{LI}{T}_i\\ & +{\beta }_{\mathit{BEH}}\mathit{BE}{H}_i+{\beta }_{\mathit{POL}}\mathit{PO}{L}_i+{\text{isin}}_i\end{array}$(1)

$\mathit{where},\mathit{EE}{R}_i=\left\{\begin{array}{c}1\mathit{EE}{R}_i^{\ast }>0\\ 0\mathit{otherwise}\end{array}\right.$

Here, $\mathit{EE}{R}_i^{\ast }$ is a latent variable for $\mathit{EE}{R}_i$, $\mathit{\beta }$ terms represent coefficients to be estimated, and ${ }_i$ denotes a stochastic error term assumed to be independent and identically distributed across the households. The model accounts for several explanatory variables: SEC denotes socio-economic characteristics, e.g. household income, household size; DW represents dwelling characteristics, e.g. building period, floor area; LIT denotes energy-related literacy and financial literacy of respondents⁵; BEH is an indicator of the energy-saving behaviour of respondents; and POL denotes policy instruments.

We use the number of renovations to create a secondary outcome variable to identify households that differ in terms of the intensity of renovation. We create three groups: households with no renovation (No EER), with 1–2 types of renovations (Some EERs), and those with three or more types of renovations (Many EERs).⁶ Given the ordinal nature of this variable, we estimate an ordered probit model of the following form:

(2) $\mathit{EERIn}{t}_i^{\ast }={\alpha }_{\mathit{SEC}}\mathit{SE}{C}_i+{\alpha }_{\mathit{DW}}\mathit{D}{W}_i+\text{break}{\alpha }_{\mathit{LIT}}\mathit{LI}{T}_i+{\alpha }_{\mathit{BEH}}\mathit{BE}{H}_i+{\alpha }_{\mathit{POL}}\mathit{PO}{L}_i+{\text{isin}}_i$(2)

$\mathit{EERIn}{t}_i^{\ast }$ is a latent continuous variable denoting the renovation intensity by household $\mathit{i}$ and $\mathit{\alpha }$ terms are coefficients to be estimated (notation for explanatory variables and ${ }_i$ is same). The probability that household $\mathit{i}$ has reached the renovation intensity group$\mathit{j}$, where $\mathit{j}$ can vary from 1 (No EER) to 3 (Many EERs), is given by:

(3) $Pr\left(\mathit{EERIn}{t}_i=\mathit{j}\right)=Pr\left(k_{\mathit{j}-1}<\mathit{EERIn}{t}_i^{\ast }<k_j\right),\mathit{where}$(3)

$-\mathrm{\infty }=k_0<k_1<\dots k_j\cdots <k_J=\mathrm{\infty },\mathit{j}\in 1,2,\dots ,\mathit{J}$

Here, $\mathit{EERIn}{t}_i$s are the ordinal values of the renovation intensity as defined above, and $k_j$s are the threshold parameters.

For both outcomes, we estimate first a panel data random effects model and another variation where we include Mundlak’s adjustment terms. All models are estimated using maximum-likelihood approach and control for the year fixed effects using year dummies.

We are especially interested to examine whether government policies aimed to promote energy-efficiency investments are influencing the investment renovation decisions. Like many European countries, Switzerland has also adopted some policy instruments to promote EER by households and several of these policies have been implemented at the cantonal level. Primarily, there are three types of measures: a) Tax deduction – availability of tax credit or deduction at the cantonal level; b) Subsidies – in the form of energy subsidy programs at the cantonal level (there is also an important federal program that promote energy saving renovation); and c) Informational measures – such as support for energy-related audit and consulting either at home or at a central location (such as the utility). Information on these policy measures were collected using external sources, and merged with the survey data at the canton and year level.⁷

We examine three policy variables: the first variable on tax deduction is dichotomous and represents whether or not a tax relief for energy-efficiency investments was available to households within a canton during a year. The second policy variable captures the per capita annual expenses of energy-related subsidy programs that includes the subsidies paid out by the cantons (including global contributions from the federal government). The third policy captures the per capita annual cantonal expenses towards provisions related to other measures, such as informational campaigns and energy-related consulting.⁸ In terms of correlations, while correlation between tax deduction and subsidy policies was very low (−0.04), we find only a weak positive correlation between the subsidy and consulting policies (0.23) and between tax deduction and consulting policies (0.15).

One limitation with our dataset is that the tax deduction policy for EERs does not vary within a canton over the study period, i.e. each of the nine canton in our data either always had a tax deduction policy in place, or did not have it at all. While we cannot change this natural experiment framework, we include a rich set of observable characteristics at the household level as explanatory variables and perform additional analysis and robustness checks.⁹ Endogeneity-related concerns, such as reverse causality with policy variables may also exist. This is less likely to be a concern with the tax deduction policy, but cannot be ruled-out in the context of the other policies. Alberini, Bigano, and Boeri (2014) point out that such programs could disproportionately attract some people who would undertake such renovations anyway. Furthermore, following Hassett and Metcalf (1995), we minimize potential omitted variable biases by specifying a rich model and by including Mundlak adjustment terms that account for unobserved heterogeneity using group means for the time-varying controls (Mundlak 1978).

Data

The dataset comes from a large survey conducted by the Centre for Energy Policy and Economics (CEPE) during 2015–2016 as part of a research project focused on energy usage by Swiss households. This dataset (hereafter, referred to as CEPE 2015 dataset) consisted of 8’378 households living primarily in urban and suburban regions of nine Swiss cantons. Comprehensive details of the survey and the data are provided in Blasch et al. (2018).¹⁰ In addition to disaggregate household level information on socio-demographics and dwelling characteristics, the CEPE 2015 dataset contains information on investments in EERs over the last five years resembling the period 2010–2014. We restrict the sample to owned, single-family households as these are likely to have more autonomy in undertaking EERs. For the final estimations, we have 1663 households that have declared that they undertook none, or some, EERs over the last five years.¹¹ About half of owned single-family households (SFHs) in the sample were built before 1970 and the average living space is 180 m² with three inhabitants. Blasch et al. (2018) compared the overall sample in the CEPE 2015 dataset to available population statistics for the Swiss population and argue that the sample resembles more to the urban and sub-urban households of Switzerland. However, it would be somewhat difficult to comment on the representativeness of our data given this unique target sample and a lack of a direct reference representative sample to compare it to. For brevity, Table 1 provides an overall summary of the cross-sectional variables whereas Table 2 presents statistics for the panel variables for the sample of households used in the empirical analysis. We note that the respondents in owned SFHs are typically older than 40 years, fall in the middle or high-income group, and about two out of five also have a university degree. In Table 2, overall around 9% households have undertaken one or more EERs and there also seems to be some variation across the years in terms of renovation intensity (EER Intensity). In terms of policy measures, most households seem to have the possibility to benefit from the tax deduction policy and there again seems to some variation across the years for the other two policy measures related to subsidy and consulting measures.

Table 1. Overview of the dataset and underlying variables.

Variable	Mean	Std. Dev.	Min.	Max.	Obs.
Number of EER (over last 5 years)	0.77	1.26	0	6	1663
Household size	2.98	1.25	1	6	1663
Size of dwelling (in sqm)	179.15	63.24	100	400	1663
Number of rooms	5.05	0.97	0	6	1663
Building built before 1940	0.26	0.44	0	1	1663
Building built in 1940–1970	0.24	0.43	0	1	1663
Building built in 1970–2000	0.35	0.48	0	1	1663
Building built in 2000–2015	0.14	0.35	0	1	1663
Household has kids/teens	0.37	0.48	0	1	1663
Household has elderly	0.34	0.47	0	1	1663
Respondent’s age < 40 years	0.07	0.25	0	1	1663
Respondent’s age between 40–60 years	0.47	0.5	0	1	1663
Respondent’s age > 60 years	0.46	0.5	0	1	1663
Household Income < 6k	0.17	0.37	0	1	1663
Household Income in 6k-12k	0.54	0.5	0	1	1663
Household Income > 12k	0.29	0.45	0	1	1663
Respondent has university degree	0.42	0.49	0	1	1663
Respondent’s partner has univ. degree	0.23	0.42	0	1	1663
Respondent’s investment literacy	0.67	0.47	0	1	1663
Respondent’s energy literacy	4.55	3.16	0	11	1663
Respondent’s energy-saving behavior	2.53	1.02	0	4	1663
Morally obliged to reduce consumption	0.73	0.44	0	1	1663
Concerned about free-riding by others	0.08	0.28	0	1	1663

Notes: This table reports the summary statistics for the sample of 1663 owned single-family households in the CEPE 2015 dataset.

Table 2. Overview of the dataset and underlying variables.

Variable (panel)	2010	2011	2012	2013	2014	2015	Overall
Atleast one EER	0.0898	0.0702	0.0971	0.120	0.0950	0.136	0.0951
	(0.286)	(0.256)	(0.296)	(0.325)	(0.293)	(0.344)	(0.293)
EER Intensity: “No EER”	0.910	0.930	0.903	0.880	0.905	0.864	0.905
	(0.286)	(0.256)	(0.296)	(0.325)	(0.293)	(0.344)	(0.293)
EER Intensity: “Some EER”	0.077	0.062	0.083	0.105	0.082	0.107	0.082
	(0.266)	(0.242)	(0.276)	(0.307)	(0.275)	(0.310)	(0.275)
EER Intensity: “Many EER”	0.013	0.008	0.014	0.015	0.013	0.030	0.013
	(0.114)	(0.089)	(0.119)	(0.121)	(0.112)	(0.167)	(0.113)
Tax deduction policy	0.878	0.884	0.885	0.883	0.883	1	0.884
	(0.328)	(0.320)	(0.320)	(0.321)	(0.322)	(0)	(0.320)
Subsidy expenses per capita	11.59	12.12	15.30	11.53	12.53	21.12	12.74
	(5.252)	(4.599)	(6.778)	(8.543)	(6.001)	(0)	(6.575)
Consulting expenses per capita	4.373	4.456	4.555	2.579	2.366	1.387	3.600
	(4.729)	(4.444)	(4.369)	(1.296)	(1.263)	(0)	(3.660)
Household size	3.089	3.024	2.969	2.920	2.907	2.738	2.976
	(1.276)	(1.256)	(1.236)	(1.228)	(1.234)	(1.120)	(1.246)
# Observations	1437	1510	1534	1552	1663	103	7799

Notes: This table reports the mean, standard deviation (in parentheses) and count for the panel variables used in the empirical analysis. The annual data covers the period from 2010 to 2014 except for one region, for which data is available from 2011 to 2015.

III. Empirical results

Table 3 reports the average marginal effects corresponding to four panel data models – Model (RE-1): probit random effects for the main outcome; Model (REM-1): probit random effects with Mundlak adjustment terms for the main outcome; Model (RE-2): ordered probit random effects for the second outcome; and Model (REM-2): ordered probit random effects with Mundlak adjustment terms for the second outcome. While we note that the estimated coefficients are quite similar across the models, for the remaining discussions, we refer to results reported in REM models (unless otherwise stated).

Table 3. Marginal effects for probit and ordered probit models for EER (CEPE 2015 dataset).

	Probit		Ordered Probit
	RE-1	REM-1	RE-2	REM-2
Tax deduction policy	0.0219*	0.0232**	0.0180**	0.0188**
	(0.0113)	(0.0116)	(0.0089)	(0.0091)
Subsidy expenses per capita	0.0007	0.0006	0.0007	0.0006
	(0.0005)	(0.0006)	(0.0004)	(0.0005)
Consulting expenses per capita	0.0016	0.0017	0.0013*	0.0015
	(0.0010)	(0.0013)	(0.0008)	(0.0011)
Household size	0.0042	−0.0068	0.0042	−0.0040
	(0.0037)	(0.0093)	(0.0029)	(0.0075)
Size of dwelling (in sqm)	−0.0001*	−0.0001*	−0.0000	−0.0000
	(0.0001)	(0.0001)	(0.0000)	(0.0000)
Number of rooms	0.0058	0.0054	0.0026	0.0023
	(0.0041)	(0.0041)	(0.0032)	(0.0032)
Building built in 1940–1970	0.0280***	0.0278***	0.0202***	0.0201***
	(0.0088)	(0.0089)	(0.0069)	(0.0070)
Building built in 1970–2000	−0.0050	−0.0051	−0.0075	−0.0078
	(0.0085)	(0.0086)	(0.0067)	(0.0068)
Building built in 2000–2015	−0.1455***	−0.1455***	−0.1148***	−0.1150***
	(0.0179)	(0.0179)	(0.0138)	(0.0139)
Household has kids/teens	−0.0103	−0.0127	−0.0082	−0.0099
	(0.0100)	(0.0102)	(0.0079)	(0.0080)
Household has elderly	−0.0108	−0.0104	−0.0085	−0.0083
	(0.0092)	(0.0092)	(0.0073)	(0.0073)
Respondent’s age < 40 years	0.0167	0.0178	0.0143	0.0150
	(0.0144)	(0.0144)	(0.0113)	(0.0113)
Respondent’s age > 60 years	−0.0134	−0.0126	−0.0116	−0.0111
	(0.0098)	(0.0098)	(0.0077)	(0.0077)
Household Income in 6k-12k	−0.0026	−0.0027	−0.0011	−0.0011
	(0.0096)	(0.0096)	(0.0076)	(0.0076)
Household Income > 12k	0.0036	0.0035	0.0035	0.0035
	(0.0117)	(0.0118)	(0.0093)	(0.0093)
Respondent has university degree	−0.0017	−0.0014	−0.0012	−0.0009
	(0.0079)	(0.0079)	(0.0062)	(0.0062)
Respondent’s partner has univ. degree	0.0170*	0.0166*	0.0150**	0.0148**
	(0.0088)	(0.0088)	(0.0069)	(0.0070)
Std(investment literacy)	−0.0020	−0.0018	−0.0013	−0.0012
	(0.0038)	(0.0039)	(0.0030)	(0.0031)
Std(energy literacy)	0.0027	0.0027	0.0026	0.0025
	(0.0035)	(0.0035)	(0.0028)	(0.0028)
Std(energy-saving behavior)	0.0046	0.0047	0.0032	0.0033
	(0.0036)	(0.0036)	(0.0028)	(0.0028)
Morally obliged to reduce consumption	0.0232***	0.0232***	0.0193***	0.0190***
	(0.0079)	(0.0080)	(0.0063)	(0.0063)
Concerned about free-riding by others	0.0259**	0.0260**	0.0205**	0.0204**
	(0.0118)	(0.0119)	(0.0093)	(0.0094)
Observations	7799	7799	7799	7799

Notes: This table reports the average marginal effects (AME) obtained for the panel data models estimated on the CEPE 2015 dataset. The main outcome variable is a dichotomous variable for whether or not any EER was undertaken (RE-1 and REM-1). The secondary outcome variable represents renovation intensity as three ordered groups (RE-2 and REM-2). For RE-2 and REM-2, the AME have been calculated using margins in Stata v13.1 and represent probability of a specified outcome level (Some EER). Coefficients on the year fixed effects and on the Mundlak terms have not been reported. Std() refers to standardized form (z-scores, with zero mean and unit standard deviation) for the three indices investment literacy, energy literacy, and energy-saving behaviour. Robust standard errors in parentheses. *, ** and *** respectively denote significance at 10%, 5% and 1% levels.

The coefficients on all three policy variables are positive and coefficients on the tax deduction policy are significant at 95% level in both REM-1 and REM-2. The subsidy-related variable has smaller coefficients and is insignificant in all models. The third policy variable is also positive but insignificant (but has a weaker significance in RE-2). It is worth noting that each of the policy considered can have different impact due to salience, public awareness, transaction costs and perceived benefit by home-owners. For example, a tax deduction policy is generally more salient as it is asked every year when filing tax returns. Whereas, subsidy expenses and consulting expenses could be linked to limited awareness, e.g. in a recent study on investment in energy-efficient vehicles, Cerruti, Daminato, and Filippini (2019) find evidence of limited awareness of energy-efficiency policies among the Swiss population. Moreover, as discussed previously, sometimes the request of subsidies can be cumbersome and time-taking.

We observe some expected findings with respect to dwelling and socio-economic variables. Compared to the buildings built before 1940 (reference category), those built in 1940–1970 are associated with a higher likelihood to undergo more retrofits whereas newly constructed buildings (after 2000) are naturally associated with a much lower likelihood. The coefficient on the highest level of household income is also positive but insignificant. While the coefficients on the literacy and energy-saving behaviour of the respondent are not significant, interestingly, level of education of the survey respondent’s partner living in the same household appears to also show an important association. Respondents who feel morally obliged to reduce their energy consumption appear to be positively associated to undertake (more) EERs. Interestingly, the same is also true for respondents who are concerned about potential free-riding behaviour by others.¹²

Overall, we conclude that while all policy measures seem to be positively associated with households undertaking EER decisions (on both types of outcomes), the role of the tax deductions seems to be more noticeable and robust. A tax deduction policy is generally more salient to the public. Owners of single-family houses belong to a high-income group and are likely to pay higher taxes. The tax deduction could be particularly interesting for home owners undertaking EERs to reduce recurring energy costs and increase their property value, and to also save on higher taxes. While we obtain similar results to Hassett and Metcalf (1995) on the positive role of tax deduction policy, the role of other policy measures appear to be somewhat inconclusive – similar to that of Alberini, Bigano, and Boeri (2014) and even to some early empirical literature (Walsh 1989; Dubin and Henson 1988).

IV. Conclusion

Increased rate of energy-saving retrofits in existing buildings will play a crucial rule in meeting energy and climate policy objectives. It is hence important to examine the determinants of household renovation decisions. While the building vintage is a relevant factor; energy-related attitudes of decision makers, as well as policy instruments are also likely to play a very important role. In particular, we find that a tax deduction policy is associated with a significant positive impact on households’ investment decisions related to energy-saving renovations, both on whether or not to undertake an energy-saving renovation, but also on the renovation intensity. We cannot completely rule out endogeneity-related concerns, so our results on the impact of the policies should be considered as suggestive evidence.

From a policy perspective, the federal and cantonal bodies should continue to promote these policies to attract consumers to undertake more energy-efficiency retrofits. Given the potential for larger reductions in energy-requirement, direct monetary policies such as tax deductions should be particularly focused on poorly insulated old buildings and those that rely on non-renewable energy sources for heating purposes. Lastly, concerted efforts from both public and private stakeholders may be necessary towards capturing systematic and comprehensive data related to policy measures at disaggregate level in order to facilitate future empirical research in this direction.

Acknowledgements

This research has been funded by the Swiss Federal Office of Energy (FOEN) under contract number SI/501886-01. Neither FOEN nor any other partners were responsible for the survey design, analysis and interpretation of the data. We would also like to thank Björn Lalin for assistance in data collection on policy measures, two anonymous reviewers for their suggestions, and researchers at CEPE, ETH Zurich for their feedback during internal discussions. All omissions and remaining errors are our responsibility.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13504851.2022.2156456.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The work was supported by the Bundesamt für Energie [SI/501886-01].

Notes

¹ The EU, as well as Switzerland, have undertaken initiatives to boost the energy-efficiency retrofits. For example, see European Commission (2021) and Swiss National Science Foundation (2021).

² For example, see European Energy Efficiency Platform (2021) and Energy Star (2021).

³ For an extensive review of different retrofit policy instruments for existing residential buildings across several countries, please refer to Zhang et al. (2021). The authors categorize around 165 retrofit policies into 4 groups of measures: direction and command, assessment and disclosure, research and service, and financial incentives, and summarize their implementation, success, and obstacles. They point out that even though buildings are ageing and several policies are available, the EER rates still remain very low across the world. They go on to identify several hurdles in successful uptake of policies, such as awareness and information issues, technical issues, financial issues, and management and other issues.

⁴ Our survey captured information on different types of EERs, namely, energy-saving retrofits of windows or doors, roof, façade, space heating system, water heating system, and other measures performed during the last five years.

⁵ For a discussion on the concept of energy literacy and its link with financial literacy and bounded rationality, see (Blasch et al. 2017; Blasch, Filippini, and Kumar 2019).

⁶ Figure A.1 in the supplementary material shows the overall number of renovations carried out by households within our sample. As basic checks, we also estimated panel count models using both Poisson regression and negative binomial approach for the number of EERs undertaken. The overall essence of the main results, important determinants etc. are similar to reported findings.

⁷ Switzerland is a federal state composed of 26 cantons that are relatively independent in defining and implementing policy measures that aim to promote energy efficient renovations. This heterogeneity in the energy policy measures can be exploited by researchers in empirical analysis. In the supplementary material, we summarize the policy information and the data sources collected as part of this exercise. While exploring external data sources, we found that there is a general lack of a systematic collection, storage and availability of information on policy measures (both current as well as historical) across different regions and cantons.

⁸ This variable also encompasses personnel and material expenses for cantonal energy policy measures, but do not include expenses towards renovation of cantonal buildings or budget for subsidy programs. Information on the three policy measures were collected from annual reports ‘Stand der Klima- und Energiepolitik der Kantone’ published by the BFE for the years 2010–2018. Table A.1 in the supplementary material presents an overview of the three policy variables across all Swiss cantons from 2010 to 2014.

⁹ We followed several propensity score matching based strategies as robustness check for the impact of the tax deduction policy. The result supports a positive role of the cantonal tax deduction policy towards EER investments. In order to restrict this analysis to more homogeneous households, we considered another specification with two adjacent cantons – one with the policy and the other without. The positive and significant role of the cantonal tax deduction policy remained robust. These results are reported in Table A.3 in the supplementary materials. However, we keep the panel data random effects model used in the main analysis as we are also able to consider time-invariant unobserved heterogeneity.

¹⁰ Different subsets of the dataset have been used in research articles focusing on some other topics related to energy efficiency and household energy use (Blasch et al. 2017; Blasch, Filippini, and Kumar 2019; Blasch et al. 2021).

¹¹ Table A.2 in the supplementary material reports summary statistics for the full CEPE 2015 dataset for reference.

¹² The dichotomous variables related to moral obligation and concern for free-riding are equal to one for respondents who agreed, or strongly agreed, to the two statements: ‘I feel morally obliged to reduce my energy consumption’. and ‘I am not willing to reduce my energy consumption if others don’t do the same’..