How to Support Residential Energy Conservation Cost-Effectively? An analysis of Public Financial Schemes in France

Chlond, Bettina; Gavard, Claire; Jeuck, Lisa Article — Published Version How to Support Residential Energy Conservation Cost- Effectively? An analysis of Public Financial Schemes in France Environmental and Resource Economics Provided in Cooperation with: Springer Nature Suggested Citation: Chlond, Bettina; Gavard, Claire; Jeuck, Lisa (2023) : How to Support Residential Energy Conservation Cost-Effectively? An analysis of Public Financial Schemes in France, Environmental and Resource Economics, ISSN 1573-1502, Springer Netherlands, Dordrecht, Vol. 85, Iss. 1, pp. 29-63, https://doi.org/10.1007/s10640-022-00754-2 This Version is available at: https://hdl.handle.net/10419/308498 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Environmental and Resource Economics (2023) 85:29–63 https://doi.org/10.1007/s10640-022-00754-2 1 3 How toSupport Residential Energy Conservation Cost‑Effectively? Ananalysis ofPublic Financial Schemes inFrance BettinaChlond1· ClaireGavard1 · LisaJeuck2 Accepted: 19 December 2022 / Published online: 6 February 2023 © The Author(s) 2023 Abstract We compare the performance of four types of support schemes aimed at improving residential energy efficiency in France: the income tax credit, a grant scheme, the reduction of the value-added tax, and the White Certificates. We use the TREMI dataset which covers close to 14,000 households that conducted conservation works. To address self-selection bias, we use a double-robust inverse probability weighting estimator. We assess the effect of the adoption of each scheme on the funding acquired, the private and total investment, and the reduction of the household energy expenses. For each scheme, we use the estimates to deduct its cost-effectiveness, the involved redistribution, and its ability to trigger additional investment in energy conservation works. We find funding from the schemes to reduce energy expenses most cost-effectively via the White Certificates. Redistribution is neutral for each of the four schemes: higher and lower income households equally benefit from them. Keywords Energy efficiency· Energy conservation· Residential retrofits· Costeffectiveness· Redistribution· Inverse probability weighting JEL Classification H23· Q58· Q84 * Claire Gavard claire.gav[email protected] Bettina Chlond [email protected] Lisa Jeuck lisa.jeuc[email protected] 1 ZEW Mannheim, L7-1, 68161Mannheim, Germany 2 Ludwig-Maximilians-University Munich, Munich, Germany 30 B.Chlond et al. 1 3 1 Introduction As global greenhouse gas emission constraints are becoming more stringent, substantial reductions must be achieved in all economic sectors. The residential sector consumes a large share of total energy. Residential users account for on average 18.4% of the total fuel consumption in OECD countries in 2017 (IEA 2019) and 25.7% in the EU (EC 2018). The energy efficiency gap, i.e. the difference between actual and optimal energy use (Jaffe and Stavins 1994), remains high in this sector1 due to reasons such as inattention of residents (Palmer and Walls 2015), the landlord-tenant dilemma (Allcott and Greenstone 2012) and high non-monetary costs of energy efficiency investments (Fowlie etal. 2015).2 In attempts to close the gap, governments offer a variety of publicly funded financial incentives, including direct subsidies and other instruments, to encourage households to conduct conservation works that improve energy efficiency and thereby also help to reduce GHG emissions.3 Evaluating the cost-effectiveness of these government policies is essential to gain insights on which support schemes achieve the highest energy reduction for each euro of public money spent. This matters for fiscal policy, especially in times of constrained public budgets. The involved redistribution between higher and lower income groups also matters. To be able to meaningfully compare instruments in an ex-post evaluation, researchers need data on different instruments that were implemented under the same administration during the same period. In this paper, we assess the performance of four types of financial schemes used to support residential energy conservation works (retrofits) that were all implemented in France between 2014 and 2016: a grant scheme for low-income households, a reduction of the value-added tax (VAT), an income tax credit and the White Certificates. The latter are certificates that French energy suppliers have to collect to a certain amount in exchange for helping energy consumers to reduce their energy consumption. Energy suppliers can then trade the certificates among each other to achieve the energy reductions set by the government. We use the TREMI2017 survey data produced by the French Environment and Energy Agency (ADEME). The dataset covers close to 14,000 observations at householdlevel and includes information on works related to energy conservation and the adoption of specific financial support schemes.4 To address potential endogeneity and self-selection bias, we employ double-robust inverse probability weighting (IPW) estimators.5 IPW uses the propensity score to compute weights based on the inverse probability of treatment. The procedure assigns high weights to observations in the control and treated group that are most comparable. 1 The European Commission plans an initiative to accelerate the annual renovation rate in the EU. The European Green Deal assumes a current average annual renovation rate of 0.4% and requests an acceleration to 1.2% (EC 2019). 2 This should be nuanced by the fact that the engineering estimates of potential savings might be too optimistic, as shown by Allcott and Greenstone 2017, Christensen etal. 2020;and Fowlie etal. 2018. 3 As examples of the amount of dedicated public spending, the Italian government invested €5.5 billion in energy efficiency in 2015, more than half of which for the residential sector, and the French government had planned to spend €2.4 billion for energy-efficient renovation of buildings in 2020 (RF 2020). For 2021, the French government had planned to spend €1.2 billion for energy efficiency works in buildings via a reduction of the value-added tax or an interest-free loan scheme (RF 2021). 4 The survey gathers information on the type of conservation works, for instance work done on the heating system or the roof. A list of work types can be found in Appendix (Table11). 5 IPW corrects the self-selection bias as long as selection can be explained by observables. 31 How toSupport Residential Energy Conservation… 1 3 The double-robust feature additionally employs predictive information of the covariates on the outcomes. As such, the estimator is robust to incorrect specification of either the propensity score or the outcome model and is consistent if only one of the models is correctly specified. We estimate the effect of adopting each scheme on the funding acquired, the private and total investment by households, and the reduction of households’ energy expenses. For each scheme, we use these estimates to evaluate the cost-effectiveness of public funds to reduce energy expenses, the ability of the schemes to induce private investment (leverage effect) and their capacity to increase total investment (additionality). We compare the four schemes along these criteria. We also analyse the distributional effects of the schemes based on the reduction in energy expenses and the funding received for lowversus high-income households. We find that the White Certificates scheme is most cost-effective, followed by the VAT reduction and the grant scheme. The income tax credit is the least cost-effec- tive. The VAT reduction triggers most additional private investment into conservation works, followed by the income tax credit and the White Certificates scheme. The grant scheme for low-income households does not trigger any additional private investment. The VAT reduction also has the highest capability to increase total investment, followed by the income tax credit and the White Certificates. The grant scheme induces least additional total investment. For each of the four schemes, households in the upper and lower half of the income distribution equally benefit in terms of funding received and reduction in energy expenses. Earlier studies set in the US context have diverse findings. While Walsh (1989) does not find any effect of a change in the income tax credit on the propensity to invest, Hassett and Metcalf (1995) find that a 10 percentage point decrease in the value-added tax on energy investment increases the propensity to invest by 24%. Estimates for the cost-effectiveness of financial support schemes to increase residential energy efficiency range between US$3.9 cent and US$47.9 cent per kWh in a heat pump rebate program in Maryland (Alberini and Towe 2015; Alberini etal. 2016). Few studies exist in the European context. Alberini and Bigano (2015) find an Italian income tax credit program to have no effect on the propensity to replace heating equipment. Blaise and Glachant (2019) assess the average impact of conservation works on energy savings in France without differentiating between support schemes. The novelty of our work in comparison with the existing literature is fourfold. First, we present a comprehensive comparison of the performance of four different financial support schemes, while previous econometric studies have commonly focused on one or two schemes. The comparison becomes possible due to the availability of data on energy efficiency works and the use of four financial schemes that were all implemented during the same time period in the same country. Second, we assess the performance of four schemes after adoption while most previous studies analyze the effect of a scheme on the uptake of conservation works. The empirical evidence on the performance after adoption is rare. Third, we compare four schemes according to the criteria of cost-effectiveness, redistribution, additionality of investment and leverage for private investment, whereas earlier studies usually focus on one dimension only, which prevents them from uncovering trade-offs between different performance dimensions. Finally, we provide new insights on the EU perspective. Previous studies that focus on the US context may not carry external validity for the EU context since they are placed in a different institutional and cultural setting. Further empirical evidence on EU member states matters. 32 B.Chlond et al. 1 3 2 Literature Review The empirical literature on financial support schemes is much focused on the US but, in the recent years, more studies on the European context were published. Numerous studies evaluate tax credits for conservation works (e.g. Alberini and Towe 2015; Hassett and Metcalf 1995; Nauleau 2014; Risch 2020; Walsh 1989); fewer studies evaluate other types of schemes, such as grants, interest-free loans and rebates (Amstalden etal. 2007; Eryzhenskiy etal. 2022; Fowlie etal. 2018). Empirical studies which compare the relative performance of schemes are scarce (e.g. Zhao etal. 2012). Policy evaluation of financial support schemes is conducted at the extensive and intensive margin; at both margins schemes can affect households’ investment behavior. First, households take the binary decision to invest into energy conservation. Schemes can affect the households’ propensity to invest, operating at the extensive margin. Second, households decide how much to invest. Here, schemes can affect the size of households’ investment, operating at the intensive margin. Previous literature focuses on the first stage decision, i.e. whether schemes encourage the uptake of conservation works. There are fewer studies that analyze the second stage decision, to what extent schemes increase the amount invested into conservation works. Two seminal papers study the effect of financial support schemes on the propensity to invest; they find contradicting evidence. Walsh (1989) assesses the effect of a change in the income tax credit rate on take-up of energy conservation works, exploiting variations between US state income tax credit rates. It is found that a change in the income tax credit rate does not affect the propensity to invest. Hassett and Metcalf (1995) assess how reduced VAT tax rates on energy investment affect the take-up of conservation works in various US states, and find that a 10 percentage point decrease in the tax increases the propensity to invest by 24%. More recent studies have likewise found ambiguous evidence. Zhao etal. (2012) compare the impact of income tax credits and interest-free loans on the propensity to invest, based on a household survey in Florida, US. They find tax credits to be more attractive than interest-free loans. The interest-free loan does not increase the propensity to invest, but the tax credit increases the rate of investing households by 12%. Alberini and Bigano (2015) find an Italian income tax credit program to have no effect on the propensity to replace heating equipment. Grösche and Vance (2008) estimate the proportion of inframarginal households that would even have invested in the absence of a financial support scheme by the German government. Around 50% of households are found to be inframarginal adopters, no hidden costs of adoption assumed. Rivers and Shiell (2016) assess the proportion of inframarginal households that adopt a natural gas furnace replacement scheme in Canada. They find that around 50% of adopters would have replaced their gas furnace even in absence of the scheme. Boomhower and Davis (2014) measure inframarginal participation in an appliance replacement program in Mexico. They estimate that at least 65% of the participating households are inframarginal and would have invested into an energyefficient appliance even without subsidy. In the French context, Nauleau (2014) investigate the effect of an income tax credit scheme6 and find a small positive effect on investment at the extensive margin for insulation measures, however at a free-riding proportion of 40 to 6 The “Crédit d’Impôt Développement Durable” (CIDD) examined by Nauleau (2014) is a predecessor of the “Crédit d’Impôt pour la Transition Énergétique” (CITE), one of the schemes studied in this paper. 33 How toSupport Residential Energy Conservation… 1 3 85%. Risch (2020) finds the same tax credit scheme to increase the impact at the extensive margin only by 1.1 percentage points, while the increase at the intensive margin is economically more significant at 22%. Eryzhenskiy etal. (2022) evaluate the effect of a zerointerest green loan program on the extensive and intensive margin investment. They find only a small effect of 3 to 4 percentage points at the extensive margin but the effect fades out in the third year after the start of the program. At the intensive margin, the amount invested increases by 9%. The cost-effectiveness of financial support schemes has been evaluated for different contexts and schemes. Gillingham etal. (2018) compare the cost-effectiveness of financial schemes from different studies on energy efficiency subsidies. Reduction costs per kWh lie in a range between US$3.9 cent and US$47.9 cent in a heat pump rebate program in Maryland (Alberini and Towe 2015; Alberini etal. 2016). Blaise and Glachant (2019) find that conservation works in France reduce the energy bill on average by only 0.64%. They express concerns about the effectiveness of financial support schemes that aim to save energy.7 Using a cost-benefit analysis, Giraudet etal. (2012) indicate that the White Certificates scheme is a cost-effective and economically efficient instrument. Financial support schemes provided by the government reallocate taxpayers’ money to recipients; distributional effects of the reallocation depend on a scheme’s design and target group. Both these features affect which socio-economic strata adopts a scheme. For instance, high-income households benefit most from income tax credits since these can deduct costs of conservation from higher tax liabilities (Neveu and Sherlock 2016). Lowerincome households preferably use grants and rebates when they are pessimistic about paying back loans. Marketing and implementation conditions can play a crucial role in determining the socio-economic background of recipients (Hoicka etal. 2014). For instance, demanding requirements of paperwork to apply for financial support schemes can be a barrier to less-educated households, and the design of information campaigns can determine to which groups a scheme reaches out (Walsh 1989). Empirical studies find recipients of financial support schemes to be a homogenous group with regressive effects on the income distribution. Using survey data, Allcott, Knittel, and Taubinsky (2015) show that the majority of households benefitting from conservation subsidies are wealthy environmentalist homeowners. Rivers and Shiell (2016) find likewise that recipients of a gas furnace replacement scheme in Canada are in large parts middle- and high-income households. 3 Institutional Background In France, the 2005 law setting the orientations of the national energy policy8 introduced the White Certificates to oblige energy suppliers help consumers to lower their energy consumption. This followed the 2002 EU Directive on energy performance of buildings (EU 2002), which introduced minimum energy performance requirements for buildings and 7 We should also mention modeling approaches such as the work by Giraudet etal. (2021) to compare the expected cost-effectiveness and distributional impacts of several support schemes in France, or the study by Charlier and Risch (2012) to compare their contribution to the reduction of energy consumption. Oikonomou etal. (2007) also conduct an ex-ante calculation of potential benefits from two packages of measures that could be used under a White certificate scheme in the Netherlands. 8 “Loi de Programmation fixant les Orientations de la Politique Énergétique” published in the Journal Officiel de la République Française (RF, (2005)) and referred to as POPE law. 34 B.Chlond et al. 1 3 energy performance certificates. In 2009, the Grenelle I law (RF 2009) set the target to renovate 800,000 social housing units in order to halve their energy consumption by 2020, to encourage the construction of low-energy consumption buildings, and aim for a 38% reduction in energy consumption in old buildings by 2020. The Grenelle II law of 2010 (RF 2010) introduced the objective to reduce energy consumption in new buildings by a factor of five by 2012. The 2010 Amendment of the 2002 EU Directive (EU 2010) stated the objective to have all new building nearly zero-energy after 2020 (after 2018 for new building occupied and owned by public authorities). In this context, the Housing Energy Renovation Plan (“Plan de Rénovation Energétique de l’Habitat”, PREH) of 2013 (RF 2013) introduced the objective to renovate 500,000 housing units per year by 20179 (including 380,000 privately owned units) and to reduce energy consumption in the housing sector by 38% by 2020. To do so, national public aids were developed, which are presented below.10 In France, households can benefit from a variety of financial support schemes for renovation works that aim to improve the energy efficiency of private dwellings. In the following, we discuss four types of at national level available schemes on which we focus our analysis: an income tax credit, a reduction of the VAT, a grant scheme and the White Certificates scheme. Further information on the financial support schemes can be found in RF and ADEME (2020). The income tax credit (“Credit d’Impôt pour la Transition Energétique”, CITE) that is offered by the French government applies to expenditures on home insulation or equipment that reduce energy consumption (see non-exhaustive list of examples in appendix). The installation costs of the works are also eligible. The work must be carried out by a company that does all the work or subcontracts some of it to another company. For most of the work, the company must be certified. The scheme allows deductions from the income tax of up to 30% of the invoice sum.11 The maximum amount of expenses that could be considered was €8,000 for a single person and €16,000 for a couple.12 Only home owners can use this tax credit scheme. Another financial instrument funded by the government is the reduction of the valueadded tax (VAT) to 5.5%. Every household is eligible to the VAT reduction as long as it conducts renovation works that target the energy efficiency of their dwellings. The VAT reduction applies to the installation, maintenance and supply of materials, appliances and equipment mentioned in a nationally defined list13 (see list in appendix) under the condition that they are compliant with technical characteristics and minimum performance criteria defined by law.14 To benefit from this reduction, households have to submit their invoice three months at the latest after the invoicing. The granted amount directly depends on the amount of private investment into energy efficiency works, funding being a fixed 9 In 2018, the “Plan gouvernemental de rénovation énergétique des bâtiments” (Governmental energy renovation plan for buildings) extended this objective of renovation to 500,000 housings per year for five more years. 10 In 2015, the “Loi de transition énergétique pour une croissance verte” (law on the energy transition of green growth) introduced the objective to achieve a level of energy performance of low-energy building standards for the entire housing stock by 2050. 11 The CITE was implemented from 2014 onwards, it replaced the CIDD. 12 These are the criteria which were in force during the time period considered for TREMI2017. 13 The list is defined in point 1 of Article 200 quater of the “Code Général des Impôts”. 14 The technical characteristics and minimum performance criteria are listed in Article 18 bis of Annex IV to the “Code Général des Impôts”. 35 How toSupport Residential Energy Conservation… 1 3 proportion of the latter; the higher the private investment, the higher the funding received from the VAT reduction is. The French housing agency (“Agence Nationale pour l’Amélioration de l’Habitat”, ANAH)15 provides the program “Habiter Mieux” to help households by supplying a direct grant of between 35% and 60% of renovation work net expenses. The maximum amount that can be allocated is €10,000 on top of which a bonus16 can be added if the energy efficiency improvement obtained is at least 25%. The grant aims at supporting households with low-income levels that live in dwellings of 15 years of age or older. The eligibility depends on the joint annual income, the number of persons in the household and whether the dwelling is located in the Paris region (Île-de-France) or not (Anah2016). The lower the household income and the higher the number of persons in the household, the higher is the maximum amount granted by ANAH. The survey data however suggests that the eligibility criteria are not strictly enforced. In the data, we see that even some households that report incomes above the eligibility thresholds of the ANAH grant receive funding from this scheme (see Fig.2 in the appendix for a detailed overview). Finally, since 2006, French energy suppliers have been obliged to collect a certain volume of “Certificats d’Economie d’Energie” (CEEs), which correspond to the so-called “White Certificates”. In exchange for assisting energy consumers to lower their energy consumption, energy suppliers receive certificates which they can trade among each other to cover the energy reductions they are required to achieve by the government. To accomplish the energy savings and to earn certificates, energy suppliers offer programs which inform households about energy consumption and savings, and financially support them to conduct conservation works. To be funded, the works have to be done by certified operators. Energy suppliers are proactive to detect consumers for whom the energy saving potential is the highest. They contact households individually, inform them about the programs they offer and encourage them to renovate their dwellings. This scheme does not only support energy efficiency of residential buildings but also of industrial and public buildings. By design, funds provided for energy conservation measures stem from the firms obtaining White Certificates, not from the public sector. The amount of funding received by households from this scheme depends on the specific program. The programs differ widely in their scope and the type of works that they support. A non-negligible share of households in our sample adopted more than one scheme. The VAT reduction has the highest absolute number of adopters in the sample and is frequently adopted in combination with one of the other three schemes. The VAT reduction seems to be a scheme “for everyone”, possibly due to a simple application process and being pointed out to households by professionals conducting the works. The grant scheme is predominantly adopted by households with lower average income due to its eligibility criteria. Households that adopt more schemes in parallel tend to be households with higher incomes and older family heads. A larger number of schemes adopted in parallel also correlates with a higher amount of acquired funding and higher private investment by households. Households that adopt more than one scheme in parallel remain in our study sample. 15 The main funding sources of ANAH are the auctioning of carbon quotas, subsidies and taxes. Additionally, energy suppliers contribute to the funding of ANAH in return for the issue of White Certificates (CEE). 16 The bonus consists of 10% of the expenses without VAT. The maximum amount that can be received is between €1,600 and €2,000. 36 B.Chlond et al. 1 3 We control for parallel adoption by including national, regional and local scheme variables in the specification IV of the treatment model (see Sect.4). 4 Material andMethods 4.1 Data We use a dataset from the TREMI survey (“Travaux de Rénovation Energétique dans les Maisons Individuelles”), conducted by KANTAR PUBLIC/TNS SOFRES, Énergies Demain and Pouget Consultants for ADEME, the French Agency for the energy transition, in spring 2017. The survey targets energy conservation works that French households conducted between 2014 and 2016. The household sample was randomly drawn from the data base of the national statistics agency (Institut National de la Statistique et des Etudes Economiques, INSEE). The questionnaire was approved by TNS SOFRES. The resulting dataset has a cross-sectional structure at the household level and contains 44,921 observations, whereof 14,081 households conducted renovation works. The survey offers a rich set of covariates, including information on the works conducted and the types of schemes used, as well as household and housing characteristics which can be used as controls. In the following, we present the variables that we employ for our analysis. Detailed summary statistics are reported in appendix A. Household characteristics include the age of the household’s reference person, the number of persons living in the household (household size), the annual household net income, the profession of the household’s reference person, the region where the household lives as well as the size of the agglomeration. The housing characteristics include the type of housing - 29,253 households in single-family houses, of which 9,964 conducted works, and 15,481 in apartments, of which 3,990 conducted works - the construction date, the status of ownership (owner versus renter) and living space of dwellings. The data also reports the funding that households acquired from the schemes, the total amount invested including public money and private investment, and the reduction of energy expenses observed in the household. In total, the dataset includes 32,876 individual renovation works. Most of the works are related to doors and windows (22.5%), walls (e.g. insulation, 18.6%), the roof (17.8%) and the heating system (17.6%). Other types of works relate to warm water (7.4%), the floors (9.3%) and ventilation (5.4%). A variable specifies the year in which all works in the household were finalized. Apart from the national schemes, households also used regional and local schemes that are only available in specific regions.17 Given the low rate of adoption of regional aids in the survey data, we only analyze the performance of national schemes. The nationally available schemes on which the analysis focuses include the grant scheme by ANAH, adopted by 7.9% of the households that conducted retrofits, the VAT reduction to 5.5%, adopted by 41.2% of the households, the income tax credit (“Crédit d’Impôt à la Transition Énergétique”, CITE), adopted by 9.8% of households, and the White Certificates scheme, adopted by 7.8% of households. Due to the design of the survey not all households were asked which schemes 17 Regional schemes reported in the dataset include “Picardie Pass renovation”, “Cheque Eco-énergie Normandie”, “Eco-chÈque de la Région Midi-Pyrénées”, “Prêt bonifié RénovLR de la Région Languedoc- Roussillon”, “AREEP de la Région Pays de la Loire”, “ISOLARIS” (Région Centre-Val de Loire) and “SEM Artee” (Région Nouvelle-Aquitaine). 43 How toSupport Residential Energy Conservation… 1 3 Table 1 Impact of scheme adoption on funding The table provides a compiled overview of coefficients of interest for all conducted regressions. The treatment model entails household and housing characteristics in specification I, household and housing characteristics as well as work types in II, household and housing characteristics, work types as well as motivation and triggers in III and household and housing characteristics, work types as well as national, regional and local schemes in IV. Table24 lists all variables. The outcome model is the same in the four specifications and entails household and housing characteristics as well as the work types. Bootstrapped standard errors are in parentheses. * p < 0.05 , ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values (I) (II) (III) (IV) Direct grant  2,756** 2,777** 2,505** 2,777** (300) (231) (223) (281) VAT reduction  671** 667*** 567** 586** (90.4) (81.6) (88.2) (75.9) Income tax credit  1,003** 1,100** 1,037** 987** (150) (152) (143) (162) White Certificates  1,199** 1,007** 836** 1,041** (172) (140) (127) (158) N 7,188 7,188 6,739 7,188 Table 2 Impact of scheme adoption on private investment The table provides a compiled overview of coefficients of interest for all conducted regressions. The treatment model entails household and housing characteristics in specification I, household and housing characteristics as well as work types in II, household and housing characteristics, work types as well as motivation and triggers in III and household and housing characteristics, work types as well as national, regional and local schemes in IV. Table24 lists all variables. The outcome model is the same in the four specifications and entails household and housing characteristics as well as the work types. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values (I) (II) (III) (IV) Direct grant   0.029 0.024 -0.026 0.011 (0.072) (0.087) (0.779) (0.095) VAT reduction   0.331** 0.326** 0.312** 0.317** (0.034) (0.038) (0.030) (0.028) Income tax credit   0.222** 0.226** 0.252** 0.189** (0.043) (0.055) (0.046) (0.049) White Certificates     0.187** 0.182** 0.165* 0.157* (0.066) (0.067) (0.068) (0.066) N 7,139 7,139 6,698 7,139 44 B.Chlond et al. 1 3 Table 3 Impact of scheme adoption on total investment The table provides a compiled overview of coefficients of interest for all conducted regressions. The treatment model entails household and housing characteristics in specification I, household and housing characteristics as well as work types in II, household and housing characteristics, work types as well as motivation and triggers in III and household and housing characteristics, work types as well as national, regional and local schemes in IV. Table24 lists all variables. The outcome model is the same in the four specifications and entails household and housing characteristics as well as the work types. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values (I) (II) (III) (IV) Direct grant 0.400** 0.402** 0.349** 0.358** (0.072) (0.069) (0.071) (0.059) VAT reduction 0.424** 0.419** 0.397** 0.418** (0.027) (0.024) (0.028) (0.024) Income tax credit 0.347** 0.349** 0.366** 0.303** (0.044) (0.036) (0.048) (0.056) White Certificates    0.378** 0.375** 0.333** 0.346** (0.048) (0.056) (0.062) (0.060) N 7,169 7,169 6,727 7,169 Table 4 Impact of scheme adoption on the reduction of energy expenses The table provides a compiled overview of coefficients of interest for all conducted regressions. The treatment model entails household and housing characteristics in specification I, household and housing characteristics as well as work types in II, household and housing characteristics, work types as well as motivation and triggers in III and household and housing characteristics, work types as well as national, regional and local schemes in IV. Table24 lists all variables. The outcome model is the same in the four specifications and entails household and housing characteristics as well as the work types. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values (I) (II) (III) (IV) Direct grant 0.162** 0.160** 0.195** 0.137** (0.026) (0.032) (0.034) (0.020) VAT reduction 0.039** (0.013) 0.040* (0.013) 0.024* (0.014) 0.041** (0.014) (0.013) (0.013) (0.014) (0.014) Income tax credit 0.050* 0.053** 0.045* 0.042* (0.023) (0.025) (0.020) (0.024) White Certificates     0.092** 0.092** 0.086** 0.087** (0.035) (0.034) (0.028) (0.035) N 6,213 6,213 5,816 6,213 45 How toSupport Residential Energy Conservation… 1 3 coefficients can then be interpreted as percentage increases in the total amount invested. The treatment effect of each scheme is estimated in a separate regression. We compile the results and report them in Table3. The effect of adoption on total investment is positive and significant for all schemes and robust across specifications. The largest increase in total investment is induced by the adoption of the VAT reduction, followed by the grant scheme. The White Certificates scheme induces less additional investment, and the income tax credit provides the lowest increase in investment. Adopting the VAT reduction increases total investment by 48.7 to 52.8%, a large increase that seems to be driven by a high increase in private investment while the induced funding for the scheme is low. The grant scheme increases total investment by between 41.8 and 49.5%, explained by the highest amount of induced funding while private investment does not increase with its adoption. Using the White Certificates increases total investment by 39.5 to 45.9%. The income tax credit increases total investment by between 35.4 and 44.2%.23 None of the four schemes analyzed appears to induce pure windfall gains to recipient households as all schemes increase the total amount invested. Reduction of Energy Expenses We estimate the impact of adopting each scheme on the reduction of energy expenses. We recode the categorical variable that provides a qualitative measure of the monetary reduction in energy expenses into a binary variable: the outcome is set to 1 for households that report a significant reduction in energy expenses after the renovation works and to 0 for households that report “a bit”, “not so much” or “not at all” for the reduction in energy expenses (summary statistics for the variable are shown in Table5). We estimate a linear probability model that provides a coefficient for the increase in the rate of households that report a significant reduction in energy expenses after renovation works due to the adoption of each scheme.24 The treatment effect of each scheme is estimated in a separate regression. We compile the results and report them in Table4. The effect of adoption is positive and significant for the four schemes across all specifications. Coefficients for the grant scheme are the highest, ranging between 0.137 and 0.195, corresponding to an increase of between 14 and 20 percentage points in the rate of households that experience a significant reduction in their energy expenses after renovation works. Lower-income households that are targeted by the grant scheme tend to live in smaller dwellings so that the same investment may improve energy efficiency more than in larger dwellings. The White Certificates scheme induces a more moderate increase of 9 percentage points. The increase induced by the income tax credit is 5 percentage points lower, and the VAT reduction has the smallest effect on the reduction rate, between 2 and 4 percentage points.25 23 Risch (2020) evaluates the effect of the predecessor scheme - the CIDD - in the period 2005-2008 and finds that the policy significantly increased renovation expenditures by 22%. The estimated effect is not directly comparable to our somewhat larger estimate because of the difference in the time period considered, changes in the policy design as well as in the overall economic environment. 24 Due to the structure of the data, some models leave out a few of the explanatory variables in the treatment model or outcome model to allow the estimations to converge. 25 Using a modeling approach, Charlier and Risch (2012) investigate the effect of the income tax credit, the VAT reduction and the grant scheme for the period 2011-2050. They provide a ranking of the expected contributions of the different schemes to the reduction in energy consumption. They find the income tax credit to perform best followed by the VAT reduction. The grant scheme is last in the ranking, contrasting with our empirical findings. 46 B.Chlond et al. 1 3 Our estimates indicate which schemes induce most funding, private and total investment as well as which schemes induce a significant reduction in energy expenses. We can however not directly compare the performance of the schemes using these results. We need to account for the facts that the schemes involve different amounts of funding, that they induce different amounts of additional private investment and that they are adopted by different groups of households. For instance, the same increase in the rate of households that significantly reduce their energy expenses may be driven by a lower amount of funding for one scheme than for another, so that the cost-effectiveness differs widely. Also, additional private investment induced by each scheme should be related to the amount of funding provided by the scheme to be able to compare the leverage effect per unit of received funding. We hence construct four metrics by means of which we compare the schemes according to their cost-effectiveness, their leverage effect for private investment, the additionality of investment via the induced funding and the redistribution between higher- and lowerincome households. 5.2 Cost‑Effectiveness, Leverage Effect, Additionality andRedistribution We construct metrics of the cost-effectiveness of funds, the ability of the received funding to trigger additional private investment, the additionality of investment via the induced funding, and the redistribution involved by each scheme. Cost-Effectiveness of Funds We compute a metric to characterize, for each scheme, the cost-effectiveness of the funding received to reduce the energy expenses. It takes into account the funding received by the households and the increase in the rate of households that experience a significant reduction in energy expenses induced by the scheme.26 We define the cost-effectiveness Φ of the received funds to reduce energy expenses under scheme A by normalizing the estimates for the reduction of energy expenses with the estimates for the funding acquired: This allows us to compare the four schemes according to their relative capacity to reduce energy expenses per €  of funding received.27 The larger the ratio, the more cost-effective the received funds are in promoting energy conservation. The computation results are presented in Table5. We find the White Certificates scheme to be the most cost-effective to reduce energy expenses: the average amount of funding received via this scheme is moderate, but the increase in the rate of households that reduce their energy expenses significantly is substantial. The income tax credit ranges below the White Certificates scheme due to the smaller increase in the reduction in energy expenses but higher funding that the schemes induces. The cost-effectiveness ranges associated with the grant scheme and VAT reduction overlap with both the lower bound of the range for the White Certificates and the upper bound of Φ=  𝛽 energy expenses reduction A  𝛽funding received A ∗ 10 000 26 We focus on these factors abstracting from other welfare-related effects such as increased comfort in the home after retrofits or administrative costs of distributing funds via the schemes. 27 Our analysis focuses on the cost-effectiveness of the received funds to reduce the energy bill of households. We do not calculate the cost-effectiveness of the private investment. 47 How toSupport Residential Energy Conservation… 1 3 the range for the income tax credit. With a high reduction in energy expenses but also high funding for the grant scheme and a low reduction in energy expenses but also low funding, both schemes’ cost-effectiveness is moderate.28 Despite criticisms of the White Certificates,29 we find this scheme to be most cost effective.30 Our result may be explained by the incentive energy suppliers have to target households whose dwellings have a high energy saving potential. Energy suppliers indeed have to purchase certificates if they do not comply with the obligation that they themselves have to save a specific amount of energy. They are hence proactive to inform customers about potential energy savings for their dwelling. For instance, some energy suppliers send customers a thermal image of their house to let them visualize the deficiencies in their dwelling’s thermal insulation. Similar to a small audit, this approach may specifically encourage a group of customers that otherwise would not be attentive to the energy saving potential for their dwelling and that have not yet invested in energy efficiency of their dwellings.31 In addition, not only is the funding spent targeted to lower cost energy savings opportunities, but also it uses mostly private funds. This reinforces even more the economic efficiency argument in favor of this scheme. Leverage Effect as Ability to Induce Private Investment To know which scheme induces the strongest increase in additional private investment per euro of funding, we divide the estimates for private investment by the estimates for the acquired funding. The larger this metric Λ , the higher the relative increase in private investment the scheme A is able to induce per euro of funding. The results of our computations are presented in Table6. We find that the leverage effect is highest for the VAT reduction. It induces the strongest increase in additional private investment per euro of funding, as the average funding received is the lowest in comparison with the other schemes and the increase in private investment is the strongest. The leverage effect for the income tax credit is lower, since this scheme only induces moderate amounts of additional private investment but higher funding. The White Certificates scheme has an even lower leverage effect as the relative increase in private investment is smaller while the funding is moderate. The grant scheme does not induce significant amounts of additional private investment; its leverage effect is 0. Additionality of Investment We characterize additionality by the relative increase in the amount invested beyond what would have been invested without adoption of a given scheme. The metric we employ is the estimates of the increase in total investment over the estimates of funding acquired Λ A=  𝛽 private investment A  𝛽funding received A ∗ 10 000 28 Using a modeling approach, Giraudet etal. (2021) study the cost-effectiveness of the VAT reduction, the income tax credit and the White Certificates. They find the VAT reduction to be more cost-effective than the income tax credit, and expect the White Certificates scheme to perform worst, contrarily to our findings. 29 See, for example, Glachant, Kahn and Lévêque (2020). 30 This finding is consistent with what Giraudet etal. (2012) conclude from their cost-benefit analysis of the White Certificates, namely that this system is a cost-effective instrument. 31 As Allcott, Knittel, and Taubinsky (2015) note, energy efficiency subsidies are generally primarily taken up by consumers that are wealthier, homeowners and more informed about energy costs so that limiting the eligibility of subsidies to specific household groups can lead to large efficiency gains. 48 B.Chlond et al. 1 3 for a scheme A as presented below. The larger this ratio Γ , the more capable the scheme is to induce additional investment for each € of funding: The results of this computation are displayed in Table7. Additional total investment per € of funding is the highest one for the VAT reduction: this scheme induces only small amounts of funding but induces the highest increase in total investment jointly with the grant scheme. The VAT reduction is followed by the income tax credit on par with the White Certificates.32 The grant scheme induces least additional total investment per euro of funding: Its adoption increases total investment substantially, but the scheme also induces the largest amount of funding. Redistribution Studies have found substantial heterogeneity in energy efficiency gaps across households so that schemes targeted at specific households (for instance, low-income households or households that have not yet participated in another program) can potentially generate larger welfare gains than general schemes (Allcott and Greenstone 2012). To assess whether the effect of the four schemes is heterogeneous across income groups, we compare the funding received and the reduction in energy expenses induced by each scheme for lower- and higher-income households separately. We divide our sample at the income median and compare redistribution to the households above and below this median. As a robustness check, we also compare the 25% richest and the 25% poorest households. To compute the welfare impact of each scheme, we need to consider not only the funding received but also the reduction in energy expenses induced by the schemes by income groups.33 We first look at the funding received by income groups. Using the double-robust IPW estimator as for the estimations on the whole sample in section 5.1, we estimate the effect of scheme adoption on funding received for the split samples of households with a net income below the median, €30,700, (lowerincome households), and households with a net income equal or higher than €30,700 Γ A=  𝛽total investment A  𝛽funding received A ∗ 10 000 Table 5 Cost-effectiveness of received funds Grant VAT reduction Income tax credit White Certificates  𝛽red.energy expenses A 0.137 –0.195 0.024 –0.041 0.042 –0.053 0.086 –0.092  𝛽funding received A 2,506 –2,777 567.9 –671.3 987.1 –1,100 836.4 –1,199 Φ 0.49 –0.78 0.36 –0.72 0.38 –0.54 0.72 –1.10 32 Using a simulation approach, Giraudet etal. (2021) study the additionality (termed leverage) induced by the VAT reduction, the income tax credit and the White Certificates. They find the VAT reduction to trigger more additional investment than the income tax credit, which is in line with our results. Contrarily to our results, they find the White Certificates scheme to perform best. 33 Giraudet and Finon (2015) suggest the White Certificates scheme actually involves an implicit energy tax component as energy suppliers increase their prices to cover the cost of meeting their target. We do not consider this channel of redistribution here. 49 How toSupport Residential Energy Conservation… 1 3 (higher-income households) for each of the four schemes.34 The results for each of the subsamples of lower- and higher-income households are reported in Table8. On average, the grant, income tax credit and White Certificates scheme induce slightly less funding to lower-income households than to higher-income households, while the VAT reduction induces slightly more funding to lower-income households. The absolute difference is largest for the income tax credit and smallest for the White Certificates scheme. However, we find that the estimated coefficients for the lower- and higher-income households do not significantly differ from each other for neither of the four schemes.35 This finding also holds true when comparing the funding received by the 25% poorest and richest households (see Appendix Table25). Additionally, we estimate the reduction in energy expenses induced by each scheme separately for the lower- and higher-income households (see Table9).36 We find that the increase in the rate of households that report a significant reduction in energy expenses is higher for the higher-income households for the grant scheme - by about 40%, and the income tax credit - by more than double. For the White Certificates scheme, the increase in the rate is very similar for higher- and lower-income households. The VAT reduction is the only scheme for which the rate of significant reduction increases much more in the lowerincome households - by about four times as much - and the increase is not significant for the higher-income households. We check the significance of the difference between lower- and higher-income households as we did for the coefficients on the funding received. Again, we find the difference between the coefficients to be statistically insignificant at conventional levels. This finding also holds true when comparing the reduction in energy expenses for the 25% richest and poorest households (see Appendix Table26). We conclude that all four schemes are neither progressive nor regressive in the sense that the welfare effect taking into account the funding received and the reduction in energy expenses does not differ significantly between higher- and lower-income households. Trade-Offs Between Schemes The metrics we defined to characterize the cost-effectiveness, additionality, as well as the leverage and redistribution effects of each scheme are interdependent and trade-offs arise between these policy targets. No scheme is found to perform best along all dimensions. The White Certificates scheme is the most cost-effective and, in addition, it uses mostly private rather than public funds. However, it is average in terms of additionality and leverage. The VAT reduction ranks average in terms of cost-effectiveness but performs best in additionality and leverage. The income tax credit ranks low in terms of cost-effective- ness and performs average for leverage and additionality. The grant scheme performs average for cost-effectiveness but it does poorly in terms of leverage and additionality. Noticeably, none of the four schemes distributes significantly more funding to either lower- or higher-income households and none significantly increases the rate of poorer households with a significant reduction in energy expenses more than the corresponding rate for richer 34 We use specification (I) which includes household and housing characteristics in the treatment model and dummies for household and housing characteristics as well as work type in the outcome model. Estimations in Section5.1 show that the coefficients in the four specifications do not deviate strongly for the outcomes funding received and reduction in energy expenses. 35 We check the significance of the difference between lower- and higher-income households with a z-test for comparison of coefficients from different regressions using the formula z = 𝛽 1 −𝛽 2 √ SE2 1+SE2 2 (Cohen et al. 2013) where 𝛽1 and 𝛽2 are the coefficients to be compared, and SE1 and SE2 are the corresponding standard errors. We find none of the differences to be statistically significant at conventional levels. 36 We use specification (I), as we do for the split sample estimations on the funding received in Table8. 50 B.Chlond et al. 1 3 ones. This means none of the schemes can be classified as either progressive or regressive. In summary, if the policy focus is on cost-effectiveness and redistribution, we would recommend the White Certificates scheme as it allows households to achieve more energy savings for a given amount of funding received, in addition to using private rather than public funds. 6 Conclusion The residential sector plays a pivotal part in efforts to reduce energy consumption and GHG emissions. At present, national governments offer a diverse spectrum of financial support schemes to encourage energy conservation works in the residential sector. Given the current public budget constraints, in particular following the COVID pandemic, the cost-effectiveness and the redistribution involved by these schemes need to be assessed. We compare the performance of the four following types of financial support schemes available in France: a VAT reduction, an income tax credit, a grant scheme and the White Certificates scheme. To account for self-selection into adoption of financial support schemes and address potential endogeneity concerns, our analysis employs double-robust IPW estimators.37 We use the TREMI2017 survey data from the French Environment and Energy Agency (ADEME). This dataset covers close to 14,000 households that conduct energy conservation works and reports information on the financial support schemes they potentially used. We estimate the effect of the adoption of each aid on the reduction of energy expenses, the amount invested and the acquired funding. We use these estimates to assess the costeffectiveness, additionality, redistribution and ability to trigger private investment. We find that the cost-effectiveness of public funds received is the highest with the White Certificates. This might be due to the active approach by energy suppliers (for example via nudging) to target households for whom the energy saving potential is the highest. Energy providers are indeed incentivized to fund high-impact works because they have an obligation to save a specific amount of energy. The cost-effectiveness is more moderate for the grant scheme and the VAT reduction. The income tax credit is least cost-effective. These other schemes require more proactivity from households than the White Certificates. An additional advantage of the White Certificates with regards to public budget constraints is that it uses mostly private rather than public funds. Additional private investment is triggered most via the VAT reduction, followed by the income tax credit and the White Certificates scheme, whereas the grant scheme does not encourage additional private investment. All four national schemes increase the total amount invested by households so that none of the schemes induces pure windfall gains. Total investment increases most with the VAT reduction, and moderately with the income tax credit and the White Certificates. The grant scheme induces least additional investment. Concerning redistribution, households in the upper and lower half of the income distribution equally benefit from each of the four schemes in terms of funding received and the reduction in energy expenses. To conclude, if the criteria for selecting a scheme are focused on cost-effectiveness and redistribution, the White Certificates seem to be the scheme policymakers should develop, 37 IPW corrects for self-selection when the selection can be explained by observables. 51 How toSupport Residential Energy Conservation… 1 3 potentially taking into account improvement suggestions made to address criticisms.38 The White Certificates are indeed found to be more cost-effective without being detrimental in terms of redistribution. Appendix A. Institutional Background Examples ofWorks That are Eligible totheIncome Tax Credit (Non‑exhaustive List) • Replacement of single glazed windows with double glazed windows; • Heating or domestic hot water equipment using wood, solar energy or other biomass; • Heat pumps other than air/air; • Closed fireplaces and indoor fireplace inserts; • Energy performance diagnosis, when not mandatory (1 per dwelling per 5 year period); • Thermal insulation materials (excluding windows or doors); • Removal of an oil tank; • Installation of a double-flow ventilation system; • Electric vehicle charging system. List ofMaterials, Appliances andEquipments That are Eligible totheVAT Reduction (According toPoint 1 ofArticle 200 Quater oftheCode Général des Impôts) • Condensing boilers; -micro-cogeneration gas boilers with an electrical output of 3 kilovolt-amperes or less per dwelling; • Thermal insulation materials for opaque or glazed walls, insulating shutters or entrance doors leading to the outside; • Insulation materials for all or part of a heat production or distribution system or hot water system; • Heating control devices; Table 6 Leverage effect for private investment Grant VAT reduction Income tax credit White Certificates  𝛽private investment A 0 0.312 –0.331 0.189 –0.252 0.157 –0.187  𝛽funding received A 2,505 –2,777 567.9 –671.3 987.0 –1,100 836.42 –1,199 Λ 0 4.6 –5.8 1.7 –2.6 1.3 –2.2 Table 7 Additionality of investment Grant VAT reduction Income tax credit White Certificates  𝛽total investment A 0.349 –0.402 0.397 –0.424 0.303 –0.366 0.333 –0.378  𝛽funding received A 2,505 –2,777 567.9 –671.26 987.0 –1,100 836.4 –1,199 Γ 1.3 –1.6 5.9 –7.5 2.8 –3.7 2.8 –4.5 38 See, for example, Glachant, Kahn and Lévêque (2020) or Langniss and Praetorius (2006). 52 B.Chlond et al. 1 3 • Energy production equipment using a renewable energy source, with the exception of electricity production equipment using the radiative energy of the sun, or heat pumps, other than air/air, whose essential purpose is the production of heat or domestic hot water; • Ground heat exchanger of geothermal heat pumps; • Equipment for the connection to a heating network, supplied mainly by renewable energy sources or by a cogeneration installation. Eligibility totheANAH Grant Scheme (See Fig.2) Table 8 Funding received by the lower- and higher-income households The table reports the coefficient of interest for the four schemes based on specification I which entails household and housing characteristics in the treatment model and household and housing characteristics as well as the work types in the outcome model. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values Grant VAT reduction Income tax credit White Certificates Low-income households 2,995** 657.7** 1,032** 919.5** < €30,700 (359.2) (126.6) (293.8) (192.1) N 3,221 3,221 3,221 3,221 High-income households 3,027** 632.8** 1,072** 933.7** ≥ €30,700 (472.7) (111.6) (178.7) (248.8) N 3,967 3,967 3,967 3,967 Table 9 Reduction in energy expenses by lower- and higher-income households The table reports the coefficient of interest for the four schemes based on specification I which entails household and housing characteristics in the treatment model and household and housing characteristics as well as the work types in the outcome model. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001. Significance levels are adjusted for multiple hypothesis testing using the False Discovery Rate sharpened q-values Grant VAT reduction Income tax credit White Certificates Low-income households 0.140** 0.065* 0.025 0.101* < €30,700 (0.030) (0.023) (0.049) (0.052) N 2,808 2,808 2,808 2,808 High-income households 0.201** 0.016* 0.056* 0.106* ≥ €30,700 (0.037) (0.018) (0.035) (0.036) N 3,405 3,405 3,405 3,405 59 How toSupport Residential Energy Conservation… 1 3 Appendix D. Redistribution: Results forHouseholds intheUpper andLower 25% oftheIncome Distribution (See Tables25, 26) Table 24 (continued) Variable I II III IV Eco-Cheque Midi-Pyrenees x Pret bonifie RenovLR Languedoc-Roussillon x Tiers-financement et prets de la SEM Artee x AREEP Pays de la Loire x Pret a taux zero ISOLARIS x Fig. 3 Distribution of propensity score estimates. Notes The figures show the distribution of propensity scores of the treated (light-grey) and control households (dark-grey) for treatment score specifications I to IV (from right to left) and for the four schemes (grant, VAT reduction, income tax credit and White Certificates from top to bottom) 60 B.Chlond et al. 1 3 Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s10640- 022- 00754-2. Acknowledgements The authors wish to thank Ulrich Wagner for advice on this research as well as Andreas Ziegler, Jay Shimshack and participants to the Econometrics Seminar at ZEW Mannheim, the 26th annual EAERE conference, and the 8th annual FAERE conference for their comments on this work. We thank Kathrin Schwörer, Ekin Demirtas and Leon Landes for excellent research assistance, as well as Pierre Gliganic for background information on the White Certificates scheme. We acknowledge financial support from the German Federal Ministry of Education and Research (BMBF) as part of the COMPLIANCE project (grant number 01LA18060). BMBF was not involved in any part of the research. The authors also thank ADEME for sharing the 2017 TREMI survey data. These original data are available upon request to ADEME. Claire Gavard joined the European Commission after this work was submitted and accepted for publication by Environmental and Resource Economics. This institution was not involved in this research. There is nothing else to declare. The opinions expressed in the paper are those of the authors. They do not reflect the opinion of ADEME, TNS-SOFRES, BMBF or any other institution. Any remaining shortcomings are the authors’ responsibility. Author Contributions Conceptualization: BC and CG; Data preparation: LJ; Methodology: BC and CG; Formal analysis and investigation: BC, CG and LJ; Writing: BC, CG and LJ; Funding acquisition: CG; Supervision: CG. Funding Open Access funding enabled and organized by Projekt DEAL. Table 25 Redistribution to households in the upper and lower tail of the income distribution The table reports the coefficient of interest for the four schemes based on specification I which entails household and housing characteristics in the treatment model and household and housing characteristics as well as the work types in the outcome model. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001 Grant VAT reduction Income tax credit White Certificates Low-income households 3,026*** 572.0*** 1,137* 800.2 < €25,000 (377.5) (163.4) (477.7) (520.8) N 1,991 1,991 1,991 1,991 High-income households 3,030* 521.6** 1,176*** 1,224*** ≥ €40,000 (1,392) (160.1) (236.8) (404.8) N 2,604 2,604 2,604 2,604 Table 26 Reduction in energy expenses of households in the upper and lower tail of the income distribution The table reports the coefficient of interest for the four schemes based on specification I which entails household and housing characteristics in the treatment model and household and housing characteristics as well as the work types in the outcome model. Bootstrapped standard errors are in parentheses. * p< 0.05, ** p<0.01, *** p< 0.001 Grant VAT reduction Income tax credit White Certificates Low-income households 0.153** 0.054 -0.025 0.118 < €25,000 (0.048) (0.032) (0.080) (0.119) N 1,760 1,760 1,760 1,760 High-income households 0.127 0.029 0.091 0.091 ≥ €40,000 (0.095) (0.021) (0.048) (0.055) N 2,236 2,236 2,236 2,236 61 How toSupport Residential Energy Conservation… 1 3 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:// creat iveco mmons. org/ licen ses/ by/4. 0/. References Alberini A, Bigano A (2015) How effective are energy-efficiency incentive programs? evidence from Italian homeowners. Energy Econ 52:S76–S85. https:// doi. org/ 10. 1016/j. eneco. 2015. 08. 021 Alberini A, Gans W, Towe C (2016) Free riding, upsizing, and energy efficiency incentives in Maryland homes. Energy J 37(1):259–281 Alberini A, Towe C (2015) Information v. energy efficiency incentives: evidence from residential electricity consumption in Maryland. Energy Econ 52:S30–S40. https:// doi. org/ 10. 1016/j. eneco. 2015. 08. 013 Allcott H, Greenstone M (2012) Is there an energy efficiency gap? J Econ Perspect 26(1):3–28. https:// doi. org/ 10. 1257/ jep. 26.1.3 Allcott H, Greenstone M (2017). Measuring the welfare effects of residential energy efficiency programs. https:// doi. org/ 10. 3386/ w23386 Allcott H, Knittel C, Taubinsky D (2015) Tagging and targeting of energy efficiency subsidies. Am Econ Rev 105(5):187–91 Amstalden RW, Kost M, Nathani C, Imboden DM (2007) Economic potential of energyefficient retrofitting in the Swiss residential building sector: the effects of policy instruments and energy price expectations. Energy Policy 35(3):1819–1829. https:// doi. org/ 10. 1016/j. enpol. 2006. 05. 018 Anah (2016) Le guide des aides Austin PC (2016) Variance estimation when using inverse probability of treatment weighting (IPTW) with survival analysis. Stat Med 35(30):5642–5655 Blaise G, Glachant M (2019) Quel est l’impact des travaux de renovation énergétique des logements sur la consummation d’énergie? Une évaluation ex post sur données de panel. La Revue de l’Énergie 646:46–60 Bodory H, Camponovo L, Huber M, Lechner M (2020) The finite sample performance of inference methods for propensity score matching and weighting estimators. J Bus Econ Stat 38(1):183–200 Boomhower J, Davis LW (2014) A credible approach for measuring inframarginal participation in energy efficiency programs. J Public Econ 113:67–79. https:// doi. org/ 10. 1016/j. jpube co. 2014. 03. 009 Busso M, DiNardo J, McCrary J (2014) New evidence on the finite sample properties of propensity score reweighting and matching estimators. Rev Econ Stat 96(5):885–897. https:// doi. org/ 10. 1162/ rest_a_ 00431 Charlier D, Risch A (2012) Evaluation of the impact of environmental public policy measures on energy consumption and greenhouse gas emissions in the French residential sector. Energy Policy 46:170–184 Christensen P, Francisco P, Myers E, Souza M (2020) Decomposing the wedge between projected and realized returns in energy efficiency programs. Rev Econ Stat 1–46 Cohen J, Cohen P, West SG, Aiken LS (2013) Applied multiple regression/correlation analysis for the behavioral sciences. Routledge EC (2018) EU energy in figures: statistical pocketbook 2018. Publications Office of the European Union, Luxembourg EC (2019) Building and renovating. The European Green Deal. https:// doi. org/ 10. 2775/ 48978 Eryzhenskiy I, Louis-GÄetan Segú M, Dastgerdi MV (2022) Zero-interest green loans and home energy retrofits: evidence from France. [hal-03585110] EU (2002) Directive 2002/91/EC of the European Parliament and of the Council of 16 Dec 2002 on the energy performance of buildings. L1/65 EU (2010) Directive 2010/31/EU of the European Parliament and of the Council of 19 May 2010 on the energy performance of buildings. L 153/13 62 B.Chlond et al. 1 3 Fowlie M, Greenstone M, Wolfram C (2015) Are the non-monetary costs of energy efficiency investments large? understanding low take-up of a free energy efficiency program. Am Econ Rev 105(5):201–04 Fowlie M, Greenstone M, Wolfram C (2018) Do energy efficiency investments deliver? evidence from the weatherization assistance program. Q J Econ 133(3):1597–1644. https:// doi. org/ 10. 1093/ qje/ qjy005 Gillingham K, Keyes A, Palmer K (2018) Advances in evaluating energy efficiency policies and programs. Annu Rev Resour Econ 10:511–532 Giraudet L-G, Bodineau L, Finon D (2012) The costs and benefits of white certificates schemes. Energy Effic 5:179–199 Giraudet L-G, Bourgeois C, Quirion P (2021) Policies for low-carbon and affordable home heating: a French outlook. Energy Policy 151:112–140 Giraudet L-G, Finon D (2015) European experiences with white certificate obligations: a critical review of existing evaluations. Econ Energy Environ Policy 4(1):113–130 Glachant M, Kahn V, Lévêque F (2020) Une analyse économique et économétrique du dispositif des Certificats d’Economies d’Energie. (tech. rep.) Study report. Mines ParisTech and PSL. https:// www. cerna. mines paris. psl. eu/ Donne es/ data17/ 1730- Synth AseCEE_ final. pdf Glynn AN, Quinn KM (2010) An introduction to the augmented inverse propensity weighted estimator. Polit Anal 18(1):36–56. https:// doi. org/ 10. 1093/ pan/ mpp036 Grösche P, Vance C (2008) Willingness-to-pay for energy conservation and free-ridership on subsidization - evidence from Germany. SSRN Electron J. https:// doi. org/ 10. 2139/ ssrn. 12809 03 Hassett KA, Metcalf GE (1995) Energy tax credits and residential conservation investment: evidence from panel data. J Public Econ 57(2):201–217. https:// doi. org/ 10. 1016/ 0047- 2727(94) 01452-t Hoicka CE, Parker P, Andrey J (2014) Residential energy efficiency retrofits: how program design affects participation and outcomes. Energy Policy 65:594–607. https:// doi. org/ 10. 1016/j. enpol. 2013. 10. 053 Huber M (2013) Identifying causal mechanisms (primarily) based on inverse probability weighting. J Appl Econom 29(6):920–943. https:// doi. org/ 10. 1002/ jae. 2341 IEA (2019) World energy outlook INSEE (2015) Revenu, niveau de vie et pauvreté en 2015 INSEE (2016) Recensement de la population 2016 INSEE (2017) Le parc de logements en France 1er Janvier 2017 Jaffe AB, Stavins RN (1994) The energy-efficiency gap: what does it mean? Energy Policy 22(10):804–810 Langniss O, Praetorius B (2006) How much market do market-based instruments create? an analysis for the case of “white’’ certificates. Energy Policy 34(2):200–211 Nauleau M-L (2014) Free-riding on tax credits for home insulation in france: an econometric assessment using panel data. Energy Econ 46:78–92 Neveu AR, Sherlock MF (2016) An evaluation of tax credits for residential energy efficiency. East Econ J 42(1):63–79. https:// doi. org/ 10. 1057/ eej. 2014. 35 Oikonomoua V, Rietbergenb M, Patel M (2007) An ex-ante evaluation of a white certificates scheme in The Netherlands: a case study for the household sector. Energy Policy 35:1147–1163. https:// doi. org/ 10. 1016/j. enpol. 2006. 02. 017 Ott L, Weber S (2022) How effective is carbon taxation on residential heating demand? a household-level analysis. Energy Policy 160:112698 Palmer K, Walls M (2015) Limited attention and the residential energy efficiency gap. Am Econ Rev 105(5):192–95 RF (2005) Loi no 2005-781 du 13 juillet 2005 de programme fixant les orientations de la politique énergétique. JORF 163/11570 RF (2009) Loi no 2009-967 du 3 août 2009 de programmation relative á la mise en oeuvre du Grenelle de l’environnement. JORF 0179/13031 RF (2010) Loi no 2010-788 du 12 juillet 2010 portant engagement national pour l’environnement. JORF 0160/12905 RF (2013) Dossier de presse. Plan d’investissement pour le logement. Les 20 mesures. Paris, le 21 mars 2013 RF (2020) Annexe au projet de loi de finances pour 2020 - Rapport évaluant l’éfficacité des dépenses fiscales en faveur du développement et de l’amélioration de l’offre de logements RF (2021) Annexe au projet de loi de finances pour 2021 - Rapport évaluant l’éfficacité des dépenses fiscales en faveur du développement et de l’amélioration de l’offre de logements RF ADEME (2020) Aides financiéres pour des travaux rénovation énergétique dans des logements existants Risch A (2020) Are environmental fiscal incentives effective in inducing energy-saving renovations? an econometric evaluation of the French energy tax credit. Energy Econ 90:104831 Rivers N, Shiell L (2016) Free-riding on energy efficiency subsidies: the case of natural gas furnaces in Canada. Energy J 37(4):239–266 63 How toSupport Residential Energy Conservation… 1 3 Robins JM, Rotnitzky A (1995) Semiparametric efficiency in multivariate regression models with missing data. J Am Stat Assoc 90(429):122–129. https:// doi. org/ 10. 1080/ 01621 459. 1995. 10476 494 Robins JM, Rotnitzky A, Zhao LP (1994) Estimation of regression coefficients when some regressors are not always observed. J Am Stat Assoc 89(427):846–866. https:// doi. org/ 10. 1080/ 01621 459. 1994. 10476 818 Scharfstein DO, Rotnitzky A, Robins JM (1999) Adjusting for nonignorable drop-out using semiparametric nonresponse models. J Am Stat Assoc 94(448):1096–1120. https:// doi. org/ 10. 1080/ 01621 459. 1999. 10473 862 Walsh MJ (1989) Energy tax credits and housing improvement. Energy Econ 11(4):275–284. https:// doi. org/ 10. 1016/ 0140- 9883(89) 90043-1 Zhao T, Bell L, Horner MW, Sulik J, Zhang J (2012) Consumer responses towards home energy financial incentives: a survey-based study. Energy Policy 47:291–297. https:// doi. org/ 10. 1016/j. enpol. 2012. 04. 070 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.