Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Determinants of households’ investment in energy efficiency and renewables: evidence from the OECD survey on household environmental behaviour and attitudes

Determinants of households’ investment in energy efficiency and renewables: evidence from the... Commons Attribution 3.0 licence. This paper provides novel evidence on the main factors behind consumer choices regarding Any further distribution of investments in energy efficiency and renewable energy technologies using the OECD Survey on this work must maintain Household Environmental Behaviour and Attitudes. The empirical analysis is based on the estimation attribution to the author(s) and the title of of binary logit regression models. Empirical results suggest that households’ propensity to invest in the work, journal citation and DOI. clean energy technologies depends mainly on home ownership, income, social context and household energy conservation practices. Indeed, home owners and high-income households are more likely to invest than renters and low-income households. In addition, environmental attitudes and beliefs, as manifest in energy conservation practices or membership in an environmental non-governmental organisation, also play a relevant role in technology adoption. Introduction economic context, such as available subsidies and energy prices (Poortinga et al 2003, Sardianou 2007, Investments in energy efficiency (EE) and renewable Mahaptra and Gustavsson 2008, Mills and energy sources (RES) are key to reduce greenhouse gas Schleich 2009, Di Maria et al 2010, Willis et al 2011, emissions and thus limit climate change. Global Mills and Schleich 2012, Michelsen and Madl- emissions have continued to grow by 44% from 2000 ener 2012, Sardianou and Genoudi 2013). to 2011 (IEA 2013) and policy action to reverse this Previous studies suggest that households’ socio- trend is urgently needed. In most OECD countries, the economic characteristics play a relevant role in tech- residential sector currently accounts for roughly 30% nology adoption. Generally a positive correlation of final energy consumption and of related CO between income and the probability of investing in emissions (IEA 2013). Further improving the effi- energy technologies is observed (Long 1993, Mills and ciency of energy use as well as boosting the energy Schleich 2010b, Sardianou and Genoudi 2013). It production from renewable sources will be essential to should be noted that this is indicative of credit con- make the necessary progress. By adopting energy straints. In the presence of perfect credit markets poor efficiency measures and renewable energy technolo- households should be able to borrow as long as their gies, households can make an important contribution investments are profitable. Generally, individuals with to reducing greenhouse gas emissions. This paper higher levels of education and those with children are seeks to make a contribution to the technology found to be more likely to adopt energy efficient tech- adoption literature by exploring the determinants of nologies (Mahaptra and Gustavsson 2008, Mills and such investments, which can include households’ Schleich 2009, Mills and Schleich 2012, Michelsen and socio-economic characteristics, characteristics of their Madlener 2012, Sardianou and Genoudi 2013). The dwellings, attitudes regarding environmental pro- impact of age on households’ probability to invest is blems or energy saving measures, households’ knowl- less clear. Some authors find that the propensity to edge about their energy consumption or the cost and adopt energy efficient or renewable technologies performance of energy conservation measures and declines with age (Mahaptra and Gustavsson 2008, renewable technologies, along with policy and Mills and Schleich 2009, Mills and Schleich 2012, © 2015 IOP Publishing Ltd Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Michelsen and Madlener 2012), while other results cost relatively easily, while assessing the total present suggest that middle-aged people are more likely to value of energy savings over the life of an investment is adopt such technologies than younger ones (Mills and a more difficult task given the uncertainty surround- Schleich 2010a, Sardianou and Genoudi 2013). ing energy savings and fluctuations in energy prices A number of studies underline the importance of (Jaffe and Stavins 1995, Hassett and Metcalf 1995). dwelling characteristics behind consumer choices. In This salience effect can lead households to give initial particular, the ownership of the primary residence is costs a higher weight than energy savings. Another an important driver of technology adoption. Evi- phenomenon that might explain the stronger empha- dence on the owner-effect is provided by Davis sis on initial costs than future energy savings often (2010) who shows that renters are significantly less observed in consumer behaviour is termed the ‘status likely to have energy-efficient refrigerators, clothes quo bias’. Kahneman and Tversky (1979) suggest that washers or dishwashers than owners, while Gilling- people normally perceive outcomes as losses and ham et al (2012) demonstrate that owner-occupied gains relative to a reference point, usually the status dwellings are more likely to be insulated than renter- quo. The authors’ empirical results suggest that peo- occupied dwellings. Other studies suggest that dwell- ple exhibit loss aversion in decision making under ings’ characteristics, such as location in a rural area uncertainty, giving much more weight to a possible and/or detachment from other houses may be indica- loss than to an equivalent uncertain gain. In the tors of space availability for investing in particular energy efficiency context, loss aversion can partly energy technologies, while the climate zone can influ- explain why consumers do not take up cost-effective ence the performance of specificenergymeasures investments, as they weight the certain initial costs (Mills and Schleich 2009,DiMaria et al 2010,Michel- (the loss) much more strongly than future uncertain sen and Madlener 2012)and thus people’spropensity benefits, even if these are in principle of an equivalent to invest. value. Technology adoption can also be affected by how Attitudes and beliefs may also play a role as a moti- households collect and process information that is vation to invest in addition to pure monetary benefits necessary to assess whether an investment is profit- and costs of an investment. Indeed, several studies find able. Whether consumers know the costs and benefits that people with strong environmental preferences are of different energy solutions, how much energy they more likely to invest in energy conservation technolo- use in their homes, or what rates of return to expect gies (Olli et al 2001, Kollmuss and Agyeman 2002,Di from energy efficiency measures is likely to affect the Maria et al 2010). Kahn (2007) underlines the impor- adoption of energy-efficient technologies (Mills and tance of living in environmentalist communities as a Schleich 2012). Attari et al (2010) provide evidence on driver for environment-friendly behaviour. He shows households’ misperceptions about energy use or sav- that individuals living in communities with a higher ings. They suggest that there is relatively little knowl- share of green party members are more likely to use edge regarding the effectiveness of different energy public transit, purchase green vehicles (e.g. hybrid), saving measures. Such limited knowledge is likely to and consume less gasoline than people in other com- determine the probability to invest in energy efficiency munities. Also Olli et al (2001) suggest that social con- and renewables. Indeed, when consumers are aware of text is important for environmental behaviour, as potential energy savings, the probability of investing in social participation correlates positively with respon- energy conservation measures increases (Scott 1997). sible environmental behaviour. However, although Di Maria et al (2010) find that the over 50% of their several studies suggest that peoples’ attitudes are pre- respondents are not aware of the potential energy sav- dictors of energy saving behaviour, often this relation- ings of compact fluorescent light bulbs compared to ship is weak, explaining only a small part of household the traditional bulbs. energy choices (Viklund 2004, Sjoberg and Engel- Regarding households’ ability to process informa- berg 2005, Di Maria et al 2010). tion, studies often refer to the concept of bounded Households’ energy use and energy conservation rationality suggesting that customers use simplified actions can also be shaped by habits, routines and decision-making processes, by using only a subset of social practices (Shove 2012). Understanding how the available information for complex decisions social practices occur might be essential to promote (Simon 1959). Research on bounded rationality sug- pro-environmental behaviour. Shove (2014) reports gests that individuals are more likely to take into the example of the ‘Cool Biz’ programme in Japan, account aspects that are easy to perceive than those which has been particularly effective in reducing that are difficult to assess, when they make an invest- energy consumption by modifying employees’ prac- ment decision (Yates and Aronson 1983). Indeed, tices of clothing. The programme establishes that gov- consumers tend to perceive the upfront investment ernment buildings would not be heated or cooled between 20 and 28 degree Celsius and male officers We refer to the economic concept of rationality, where individuals would be encouraged to remove jackets and ties in the are perfectly informed and use all the available information correctly to make optimal decisions. summer, while wearing more in the winter. 2 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Finally, energy prices and a favourable policy con- presents and discusses the empirical results. The final text should also affect household technology adoption. section concludes. In particular, recent studies underline the relevance of urban climate governance, that is the policies and measures undertaken by local governments for energy Data description investment to occur (Bulkeley and Broto 2013). Those policies usually include behaviour change initiatives The survey data were collected through an online aiming at promoting environmental friendly beha- questionnaire, the second of its kind, which was viour; demonstration activities promoting technologi- carried out in early 2011, while the first was launched cal innovations; transition programmes supporting in 2008. The more recent survey, which is the basis for social practices and technological systems to lower the analysis in this paper, collects data from a sample energy consumption and advocacy initiatives attempt- of more than 12 000 respondents, approximately 1000 ing to change public and political approaches to households for each country: Australia (shorthand: improve carbon emission reduction (McGuirk AUS), Canada (shorthand: CAN), Chile (shorthand: et al 2014). CHL), France (shorthand: FRA), Israel (shorthand: This paper provides novel evidence on the main ISR), Japan (shorthand: JPN), Korea (shorthand: factors behind consumer choices regarding the adop- KOR), the Netherlands (shorthand: NLD), Spain tion of energy efficiency and renewable energy tech- (shorthand: ESP), Sweden (shorthand: SWE) and nologies. It is based on the OECD Survey on Switzerland (shorthand: CHE). Household Environmental Behaviour and Attitudes For representativeness, the sample was stratified in comprising household data from 11 OECD countries. each country according to different parameters: age, Results suggest that households’ propensity to invest gender, region and socio-economic groups . Age was in clean energy technologies depends on home owner- stratified using the following groups: 18 to 24, 25 to 34, ship, income, as well as attitudes and beliefs. Indeed, 35 to 44, 45 to 54 and 55 to 69. Gender was approxi- home owners and high-income households are more mately half male and female for all countries. Region likely to invest than renters and low-income house- was stratified and quotas created using three to five holds. Households, who demonstrate environmental regions. For income stratification, households’ after- consciousness, e.g. through membership in a non- tax income quintiles were estimated for each country, governmental organisation (NGOs), in particular then responses from the survey income question were when it is environmental, or through regular energy used to fill the quotas. When quotas were filled, conservations actions, are more likely to invest than respondents with these characteristics were stopped others. from completing the questionnaire. The target The contribution to the literature on households’ respondent was between 18 and 70 years of age and technology adoption behaviour is threefold. First, the had influence on household purchasing decisions and unique dataset underlying this study includes data expenditures. Despite rigorous efforts regarding strati- from 11 OECD countries. To our knowledge, this is fication and quota sampling, it is important to the first cross-country analysis of household invest- acknowledge that there may be some respondent char- ment in energy technologies for countries across the acteristics that were not observed and which correlate OECD. Mills and Schleich (2012) analysed the adop- with internet use. This correlation of unobserved tion of residential energy technologies at the Eur- characteristics could introduce a selection bias in the opean level, while other studies focused on sample. More details on the questionnaire design, technology adoption for single countries. Second, the respondent targeting and quota sampling are provided data allow us to account for a rich set of variables, in OECD (2013), annex B. including respondents’ beliefs, attitudes and beha- viour regarding the environment and their knowledge The OECD ran ‘Call for Tender’ to select a survey service provider about their energy use and spending, in addition to specialised in the implementation of large international web-surveys using online consumer panels in different countries. Global Market more commonly investigated factors, such as house- Insite (GMI) was selected to run the survey and respondents were holds’ socio-economic characteristics, dwelling char- recruited from GMI’s in-country panels. In some countries, GMI acteristics and economic variables. This provides partnered with in-country firms with their own panels in order to further increase panel size. All partners were selected on the basis of novel insights into technology adoption behaviour. quality of panel management. Specifically GMI and its partners Third, this study covers seven different technologies, managed their respondents in line with ESOMAR 26, which is a including energy efficiency measures and renewable standard for transparency and accountability in the use of respon- dent panels for web-based survey research. To limit the risk of energy technologies, revealing differences and simila- recruiting ‘professional respondents’, GMI only permitted panellists rities regarding the determinants of investment in to answer and receive compensation for up to five questionnaires these technologies. per year. Moreover, potential respondents who started the ques- tionnaire were asked whether they met the screening criteria (living The paper is structured as follows. The following in non-institutional settings and influential in household financial section presents the data, while the third section pre- decisions). If they did not meet the criteria, they were screened out sents the econometric model. The fourth section of the sample. 3 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 1. Rates of Technology adoption across countries. Appl Bulb Pump Solar Thrm Heat Windows Country Mean Mean Mean Mean Mean Mean Mean Australia 0.69 0.91 0.03 0.20 0.58 0.15 0.13 Canada 0.67 0.87 0.04 0.04 0.38 0.65 0.51 Chile 0.41 0.95 0.01 0.02 0.31 0.06 0.14 France 0.74 0.86 0.05 0.06 0.45 0.44 0.59 Israel 0.59 0.84 0.03 0.67 0.20 0.11 0.13 Japan 0.48 0.48 0.01 0.04 0.20 0.07 0.19 Korea 0.69 0.63 0.03 0.07 0.38 0.59 0.49 Netherlands 0.61 0.89 0.02 0.04 0.49 0.48 0.73 Spain 0.74 0.91 0.02 0.06 0.21 0.47 0.54 Sweden 0.62 0.87 0.16 0.04 0.29 0.34 0.39 Switzerland 0.62 0.79 0.07 0.05 0.37 0.43 0.49 Total 0.62 0.82 0.04 0.11 0.34 0.33 0.38 The aim of this study is to investigate which factors initial investment cost and easy implementation. However, technology adoption varies significantly might drive household decision making when it comes to the adoption of clean energy technologies. The sur- across countries. Israel is the only country showing a high rate of adoption for solar panels (67%), while on vey data provides a good basis for this, as households average 11% of households have invested in solar tech- were asked whether they installed or bought appli- nologies. In the Netherlands relatively large shares of ances that received a top rating in terms of energy effi- households seem to have invested in thermal insula- ciency between 2001 and 2011. The shorthand for the tion, heat thermostats and energy-efficient windows. corresponding variable used in this paper is ‘Appl’; the Australia, as well, shows a high rate of adoption for variable takes a value of 1 for households who invested thermal insulation (58%), while Canadian households and zero for households, who could have invested, but are particularly likely to invest in heat thermostats decided against it. The same variable is constructed for (65%). In general, Japanese and Chilean households low-energy light bulbs (shorthand: Bulb), energy-effi- invest relatively infrequently in most of the technolo- cient windows (double or triple glazing, shorthand: gies considered in this study, except for energy effi- Windows), thermal insulation of walls or the roof cient appliances and low-energy light bulbs. (shorthand: Thrm), heat thermostats (shorthand: Based on the empirical literature and data avail- Heat), solar panels for electricity or hot water (short- ability, factors that might influence the decision to hand: Solar) or ground source heat pumps (short- invest in energy technology have been grouped in four hand: Pump). The survey also includes data regarding different categories: (1) socio-economic character- wind turbine investments. However given the limited istics of households; (2) the characteristics of their number of investors, namely 158 households, those dwelling; (3) households’ attitudes, knowledge and data were not included in the analysis. To study the behaviour regarding the environment; (4) house- determinants of household investment decisions only holds’ knowledge about their energy spending those households were considered who could in prin- and use. ciple have invested, while those who declared that Socio-economic variables available in the house- their house was already equipped or that, as renters, hold data set include the respondent’s age (Age), gen- they were not allowed to invest were not included in der (Female), household size, the number of years of the analysis. Table 1 shows descriptive statistics for the education after high school (Education), annual net adoption of different technologies. household income (Log_Income), the educational Among the investments considered, low-energy status of the household head, operationalised as a light bulbs were particularly frequently adopted across dummy variable for household heads who are highly countries, with more than 80% of households stating qualified professionals (Prime-earner is high skilled that they had bought such bulbs over the last ten years. worker). There is also a dummy variable that takes a Energy efficient appliances were also relatively fre- value of 1 for households stating that they cannot cope quently adopted, by more than 62% of households, with their current income (NoCope). Descriptive sta- while ground source heat pumps were adopted by only tistics are shown in table 2. a small minority of households, 4% across all coun- Households’ average annual net income is tries. Those numbers suggest that technology adop- approximately 37 868 dollars with considerable differ- tion is more likely for investments with relatively low ences in means across countries. Households living in Chile declared the lowest average annual income Heat thermostat is a device that establishes and maintains a desired (13 585 dollars), while households resident in Switzer- temperature automatically or signals a change in temperature for manual adjustment. land declared the highest level of annual average 4 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 2. Socio-economic characteristics of households. Household Prime-earner is high a a a Age Education Income Size Female NoCope skilled worker Country Mean St Mean St Mean St Mean St Mean Mean Mean Australia 42.20 14.16 3.31 3.47 48 700 27 933 2.90 1.48 0.51 0.39 0.19 Canada 43.59 14.18 3.13 2.94 42 026 26 803 2.51 1.18 0.51 0.37 0.16 Chile 37.41 12.41 4.36 3.04 13 585 10 387 3.84 1.56 0.52 0.44 0.36 France 43.18 14.07 2.64 2.36 38 157 17 697 2.74 1.17 0.51 0.44 0.12 Israel 38.3 13.24 3.95 3.45 26 562 15 329 3.63 1.65 0.55 0.43 0.32 Japan 43.67 13.83 4.90 4.26 48 394 28 702 2.99 1.49 0.49 0.29 0.08 Korea 38.53 11.66 3.26 2.38 27 012 13 892 3.49 1.32 0.50 0.33 0.13 Netherlands 45.18 13.72 4.14 3.16 38 708 16 953 2.63 1.18 0.50 0.24 0.21 Spain 41.72 12.77 3.66 3.07 29 360 16 337 2.99 1.11 0.49 0.37 0.25 Sweden 43.63 14.45 2.39 2.51 41 575 19 181 2.39 1.17 0.48 0.34 0.16 Switzerland 44.21 14.14 2.74 2.66 62 278 29 666 2.67 1.37 0.52 0.36 0.09 Total 42.01 13.77 3.50 3.15 37 868 24 681 2.98 1.42 0.51 0.36 0.19 For dummy variables, standard deviation is not computed. Table 3. Characteristics of dwellings. detached house (54%). Higher rates of ownership are a a a House Tenure Owner Rural observed in Spain (70%), Korea (70%) and the Neth- erlands (68%), while relatively many households live Country Mean Mean Sd Mean Mean in a detached house in Australia (83%), Chile (77%) Australia 0.83 9.36 11.27 0.62 0.20 and the Netherlands (75%). On average, households Canada 0.65 10.72 12.43 0.63 0.27 have lived for approximately 13 years in their primary Chile 0.77 12.95 13.94 0.65 0.14 residence, although average tenure is longer in Japan, France 0.61 12.84 13.81 0.61 0.54 around 18 years. Israel 0.32 15.10 15.62 0.67 0.20 A number of variables reflect respondents’ beliefs, Japan 0.60 18.83 16.70 0.58 0.31 attitudes and behaviours regarding the environment. Korea 0.30 8.63 9.10 0.70 0.07 This includes a dummy variable for households that Netherlands 0.75 15.99 14.81 0.68 0.53 Spain 0.26 13.76 12.97 0.80 0.38 participate in a non-governmental organisation Sweden 0.47 10.74 12.34 0.60 0.47 (NGO) and another one for those that are specifically Switzerland 0.36 11.70 12.05 0.38 0.61 in an environmental NGO (Env NGO). There is a Total 0.54 12.86 13.68 0.63 0.34 dummy variable for people who rated the environ- ment as the most pressing concern (Env_top_cncrn) For dummy variables, standard deviation is not computed. and another one for those who instead rated the econ- omy as the most pressing concern (Eco_top_cncrn). income (62 278 dollars). 36% of households stated Another dummy variable is used for those respon- that their salary was not enough to cover their needs, dents who were able to identify the causes of climate and difficulty to cope with income is particularly an change correctly (Understand_CC). issue in Chile (44%), France (44%) and Israel (43%). Respondents were asked questions regarding their It is quite surprising that Chile and Israel are the two willingness to make sacrifices to protect the environ- main countries with a higher percentage of profes- ment, their assessment of the need to do so and the sionals as household heads, 36% and 32%, respec- role of technology in solving environmental problems. tively, while France showed one of the lowest shares Depending on their answers to those questions house- (12%). Those data could partly result from French holds were grouped in three clusters : i) the envir- respondents’ difficulty to classify their occupation onmentally motivated, who are willing to make according to the categories they were given, as they sacrifices in their lifestyle to solve environmental pro- particularly frequently classified their occupation as blems (Altruists), ii) environmental sceptics who are ‘other’. The average length of education after high not willing to make much effort to solve environ- school is 3.5 years, suggesting that a number of mental problems, which they believe are often exag- respondents went to university. gerated (Sceptics), and iii) a group of technological The survey includes some characteristics of dwell- optimists who believe that environmental problems ings, such as home-ownership versus rental (Owner), dwelling type (House), years lived in the primary resi- To uncover these attitudinal profiles the latent class method (LCA) dence (Tenure) and whether households live in a rural is used. LCA is a statistical method for identifying unmeasured class membership among subjects using categorical and/or continuous area (Rural). Table 3 lists the variables used. observed variables. A description and demonstration of LCA in the The majority of respondents (63%) own their resi- context of environmental attitudes can be found in Morey dence and more than half of households live in a et al (2006). 5 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 4. Respondents’ beliefs, attitudes and behaviours regarding the environment. Env Top Eco Top Understand Cost a a a a a b Green Growthers Altruist Sceptics NGO Env NGO Concern Concern CC Bias Country Mean Sd Mean Sd Mean Sd Mean Mean Mean Mean Mean Mean Australia 0.10 0.30 0.42 0.49 0.44 0.50 0.52 0.09 0.18 0.32 0.27 0.35 Canada 0.14 0.35 0.46 0.50 0.37 0.48 0.53 0.10 0.13 0.36 0.30 0.31 Chile 0.26 0.44 0.54 0.50 0.19 0.39 0.69 0.17 0.19 0.27 0.29 0.45 France 0.10 0.30 0.56 0.49 0.33 0.47 0.44 0.08 0.11 0.44 0.24 0.25 Israel 0.12 0.32 0.64 0.48 0.23 0.42 0.50 0.12 0.09 0.21 0.15 0.39 Japan 0.19 0.39 0.31 0.46 0.46 0.50 0.32 0.03 0.17 0.49 0.26 0.36 Korea 0.25 0.44 0.38 0.48 0.36 0.48 0.42 0.05 0.28 0.37 0.21 0.33 Netherlands 0.24 0.42 0.29 0.45 0.43 0.49 0.52 0.12 0.11 0.27 0.32 0.32 Spain 0.22 0.42 0.37 0.48 0.38 0.49 0.57 0.10 0.04 0.62 0.21 0.41 Sweden 0.09 0.29 0.55 0.49 0.34 0.479 0.56 0.11 0.25 0.25 0.44 0.44 Switzerland 0.15 0.36 0.48 0.50 0.36 0.48 0.63 0.20 0.22 0.26 0.33 0.40 Total 0.17 0.38 0.45 0.50 0.35 0.48 0.52 0.11 0.16 0.35 0.27 0.36 For dummy variables, standard deviation is not computed. This variable considers that the rating given to initial investment cost exceeds the rating for future energy prices by three points on the scale. are real and technological innovations are key to sol- with some country variation. This share is 44% of ving them (Green Growthers) (OECD 2013). Respon- households in Sweden and only around 15% in Israel. dents were grouped according to their agreement on At the same time, Swedish households are more likely seven statements: (1) policies introduced by govern- to make sacrifices in their lifestyle to solve environ- ment to address environmental issues should not cost mental problems. On the other hand, Dutch house- me extra money, (2) I am willing to make compro- holds are the less likely to sacrifice their lifestyle for the mises in my current lifestyle for the benefit of the environment, although 32% of them are aware about environment, (3) protecting the environment is a the causes of climate change. Quite surprisingly, Israel shows the lowest level of awareness regarding the cau- means of stimulating economic growth, (4) Environ- mental issues will be resolved in any case through tech- ses of climate change (15%), but at the same time this nological progress, (5) environmental impacts are is the country with the highest percentage of respon- frequently overstated, (6) I am not willing to do any- dents who are environmental ‘Altruist’ (64%). thing about the environment if others do not do the A variable is constructed to capture a bias, in that a same, (7) environmental issues should be dealt with much larger weight is given to initial investment costs than to higher energy prices, although the impact on primarily by future generations. Table 4 summarises households’ propensity to invest should be the same the variables related to social context and environ- mental behaviour. for investment cost reductions and energy price The percentage of respondents who believe that increases that have an equivalent present value. environmental problems are real and express a will- Households were asked to rate different reasons that ingness to make compromises in their lifestyle to solve would induce them to invest in energy efficiency or them is 45%, although with some country variation, as change their behaviour to save more energy on a scale of 0 to 10. A variable named ‘Cost bias’ identifies 64% of respondents are ‘Altruists’ in Israel and around 55% in France and Sweden. On the other hand, on households giving a significantly higher rating to lower average across countries 35% of respondents are scep- initial investment costs than to higher energy prices. When the rating given to initial investment cost tical about environmental problems. Japan showed the highest level of scepticism (46%), while Chile exceeds the rating for future energy prices by three showed the lowest level (19%). This mirrors the share points on the scale, respondents are considered as hav- of environmental ‘Altruists’ in those two countries, ing a bias towards initial investment costs and the dummy variable takes a value of 1. As a sensitivity test, 31% in Japan and 54% in Chile. In most countries, more than 50% of respondents the variable is also constructed considering a differ- are engaged in some non-governmental organisation ence of 4 points on the scale. On average, 36% of respondents give a rating to initial investment costs (NGO). Only in Japan is this share much lower, just above 30%. On average across countries around 10% that is 3 points higher than the rating for energy prices, of respondents are engaged in an environmental 30% give a rating to initial investment costs that is 4 points higher. The highest rate is observed for Chile, NGO, but both in Japan and Korea this share is much lower. which is also the country with the lowest average Less than one third of respondents (27%) seemed annual income. This could suggest that financial con- to understand the causes of climate change, although straints might partly explain why households give a 6 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 5. Households’ knowledge about their energy spending and use. a a a Metered Ebill_known KWatt_known Behaviour Index Country Mean Mean Mean Mean Sd Australia 0.91 0.66 0.14 7.91 1.66 Canada 0.81 0.53 0.12 7.10 1.77 Chile 0.91 0.71 0.18 8.32 1.63 France 0.94 0.66 0.15 7.95 1.59 Israel 0.89 0.65 0.13 7.64 1.69 Japan 0.97 0.58 0.21 7.08 1.86 Korea 0.96 0.12 0.26 7.80 1.70 Netherlands 0.91 0.38 0.26 7.06 1.75 Spain 0.93 0.63 0.13 8.39 1.45 Sweden 0.90 0.56 0.37 5.55 1.84 Switzerland 0.90 0.63 0.15 6.82 1.77 Total 0.91 0.55 0.19 7.43 1.86 For dummy variables, standard deviation is not computed. stronger weight to investment costs than to future When interpreting these results, it is important to energy prices. However, a high percentage of biased keep in mind that response bias might occur. The wording of questions, the response scale (especially for consumers is also observed in Switzerland, the country attitudinal questions), context and data collection with the highest level of average annual income. techniques can all affect the way responses are pro- The data also capture households’ knowledge vided. In particular, even the slightest suggestion in the about their energy spending and use. A large majority way a question regarding opinions and attitudes is for- of respondents, 91% on average across countries, sta- mulated can potentially lead the respondent toward a ted that their energy consumption is metered. It is not particular answer. For instance, some results concern- very common for households to be informed about ing households’ rationality could reflect the respon- their energy bills and use, though. Respondents were dents’ difficulty to interpret correctly the questions asked to get hold of their energy bills before answering themselves. In addition, self-report bias can occur. the survey, but only about 55% were able to provide Generally, survey participants have the tendency to information about their energy spending on average respond in a way that makes them look as good as pos- across countries. Even fewer households were able to sible, or socially desirable, and this could have a bear- provide information about their energy consumption, ing on the survey findings (King and Bruner 2000, less than 19% on average across countries. The data Donaldson and Grant-Vallone 2002). Another source show an unusual result for Korean households, who of bias can be strategic misrepresentation. When seem better informed about their energy consumption respondents expect a possible connection between in volumes than about their energy spending. their response and some economic outcome in which Regarding energy use, the ‘behaviour index’ vari- they have an interest, they may have strategic incen- able captures whether respondents perform certain tives to misrepresent information. For instance, when energy conservation actions regularly, such as turning households are asked whether high energy prices off the lights when leaving the room, cutting down on would induce them to invest in energy efficiency, they heating/air conditioning to limit energy consumption, may understate their reaction to energy prices if they believe that their response could lead to an increase in running full loads when using the washing machines, future energy prices, perhaps because they think that washing clothes using cold rather than warm/hot survey results might induce the government to raise water, switching off the standby mode of appliances energy taxes. and air dry laundry rather than using a clothes dryer. The behaviour index ranges from 0 to 10, where higher values indicate that households perform several of Method these actions regularly. The data suggest that house- holds perform quite regularly energy conservation Households’ investment in energy efficiency and actions and on average across countries the behaviour renewables is investigated within a discrete choice modelling framework. For each investment good i index takes values around 7. Lower values of the index studied in this paper, households’ investment is are observed in Sweden (5.55) and Switzerland (6.82), modelled as: while higher values are observed in Chile (8.32) and Spain (8.39). Table 5 summarises the variables related yx=+βϵ (1) ii to household’s knowledge about their energy spending and use. 7 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt where y is a latent variable that captures households’ evaluate marginal effects this way and we follow this preference for technology i, namely the difference practice in most cases. Yet, as marginal effects may be between the marginal benefit and the marginal cost of very different at data points that are different from the adopting this good. X is a vector of explanatory i sample mean, it can be useful to examine marginal variables (e.g. socio-economic characteristics, dwell- effects across a range of values for some explanatory ings’ characteristics, households’ attitudes, knowledge variables, which we explore for income. In this exer- and behaviour, and household’s energy use), β is the cise dummy variables take the value that is most fre- parameter vector to be estimated and ϵ is the error quently observed in the sample, while only continuous term. While preferences cannot be observed directly, variables, except income, are evaluated at the sample the decision to adopt technology i can be observed and mean. The logit model is estimated with country-level it is modelled in line with the following decision rule: fixed effects that capture differences in policy and eco- nomic context, income levels and other country-spe- yi=<00 fy i i cific circumstances. Sweden is the base country. yi=⩾10 fy (2) The Bayesian model averaging (BMA) method is used to determine the best model specification. In That is, a household invests in good i y = 1 if ( ) absence of a theoretical model, BMA offers a systema- the marginal benefit of adopting this good is larger tic method for analysing specification uncertainty and than or equal to the marginal cost, otherwise it does checking the robustness of results to alternative model not invest(0 y = ). The probability of households’ specifications (Raftery 1995). For each tested explana- investing in good i is modelled as follows: tory variable, BMA provides the probability that this exp x β () i variable is included in the true model. This is calcu- Py== 1 x () lated on the basis of weights assigned to each tested 1e + xp x β () i models. The BMA method selects the ‘best’ model (the =Λβ x (3) () one with highest posterior probability) based on all possible combinations of the explanatory variables. In where Λ denotes the logistic cumulative distribution this paper, BMA also helps to deal with collinearity function. issues in the canonical regressions, which include all Given the non-linearity of the logit model, mar- explanatory variables. In particular, collinearity occurs ginal effects have to be calculated from the underlying with attitudinal variables, such as altruists, green estimates. For continuous variables, the marginal growthers and sceptics, while in the model selected effect measures the change in the predicted probability with the BMA method those variables are never inclu- of observing that a household invests (y = 1) associated ded all together. with changes in the explanatory variables (X ) that are infinitesimally small. For dummy variables, the mar- ginal effect measures how the predicted probability of Results and discussion observing that a household invests (y = 1) changes as the dummy variables change from 0 to 1. It should be This section discusses results from the preferred model noted that marginal effects do not have a clear inter- selected with the BMA method (tables 6 and 7). Some pretation for ordinal variables, such as the behaviour variables which have been never included in the index, as differences between different levels of this preferred model are not reported in the tables. These variable are not meaningful. Only the sign of the mar- variables include the occupation of the household ginal effect can be interpreted in this case. head and an index variable for those respondents who In this study, marginal effects are evaluated at the rated the environment or economy as the most sample means of the independent variables. In parti- pressing concern. Results from the second and third cular, marginal effects for continuous variables are best model, along with canonical regressions that computed as follows: contain all available explanatory variables, can be made available upon request. ∂= Pryx 1 () i Results suggest that socio-economic character- ⎡ ⎤ =  Λβxx1( − Λ β  β4) ()ii( ) ⎣ ⎦ istics of households partly explain investment in ∂x energy efficiency and renewables. The age of the respondent appears to be a relevant variable for most For dummy variables, the marginal effect is the difference between the predicted probability to invest of the technologies analysed. Investments in light bulbs, heat thermostats, thermal insulation and with the dummy taking a value of 1 and of zero, when all other independent variables, including dummy energy-efficient windows depend positively on age, while the probability to choose heat pumps decreases variables, are evaluated at the sample mean. While this is not very intuitive, considering that dummy variables with age. Earlier studies confirm that the probability of investing declines with age for innovative heating sys- for actual observations take either the value 0 or 1, but never the sample mean—as an example, nobody can tems, such as heat pumps, (Mahaptra and Gus- possibly be 52% female—it is common practice to tavsson 2008, Michelsen and Madlener 2012), while it 8 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 6. Bayesian Model Averaging Estimates. Logit regressions I. Dependent variables: investments in energy-efficient appliances, light bulbs and heat thermostat. Energy-efficient appliances Light bulbs Heat thermostat Variables Coefficients Marginal effects Coefficients Marginal effects Coefficients Marginal effects Age 0.0140*** 0.001 71*** 0.0115*** 0.002 29*** (0.002 26) (0.000 274) (0.002 34) (0.000 465) HHsize 0.135*** 0.0165*** (0.0249) (0.003 02) Education Log_Income 0.362*** 0.0833*** 0.0154*** 0.003 06*** (0.0425) (0.009 78) (0.0027) (0.000 538) NoCope Owner 0.336*** 0.0783*** 0.151** 0.0186** 0.372*** 0.0714*** (0.0518) (0.0122) (0.0628) (0.007 89) (0.073) (0.0134) House 0.295*** 0.0363*** 0.167** 0.0330** (0.0656) (0.008 14) (0.0694) (0.0136) Tenure −0.0502** −0.006 11** −0.0908*** −0.0181*** (0.0247) (0.003 01) (0.0258) (0.005 13) Rural 0.0206 0.002 50 0.150** 0.0301** (0.0651) (0.007 89) (0.0654) (0.0133) Green_Growther Altruist Sceptics −0.200*** −0.0248*** (0.0566) (0.007 19) NGO 0.345*** 0.0797*** 0.416*** 0.0511*** 0.270*** 0.0535*** (0.0485) (0.0112) (0.0563) (0.006 96) (0.0584) (0.0115) ENV_NGO Understand_CC Behaviour Index 0.140*** 0.0322*** 0.161*** 0.0196*** 0.0916*** 0.0182*** (0.0139) (0.003 19) (0.0154) (0.001 86) (0.0167) (0.003 32) KWatt_know 0.316*** 0.0708*** −0.007 62 −0.00 151 (0.0604) (0.0131) (0.0743) (0.0148) Ebill_know 0.189*** 0.0232*** −0.0938 −0.0188 (0.0592) (0.007 34) (0.0655) (0.0132) Metered 0.354*** 0.0845*** (0.0939) (0.023) Cost_Bias −0.125** −0.0247** (0.0599) (0.0117) AUS 0.0831 0.0189 −0.0286 −0.003 51 −1.214*** −0.185*** (0.121) (0.0273) (0.165) (0.0205) (0.153) (0.0165) CAN 0.158 0.0357 −0.283** −0.0374* 1.157*** 0.266*** (0.117) (0.0259) (0.141) (0.0202) (0.129) (0.0316) CHE −0.109 −0.0255 −0.755*** −0.114*** 0.263* 0.0550* (0.12) (0.0283) (0.132) (0.0237) (0.136) (0.0296) CHL −0.690*** −0.167*** 0.365** 0.0400** −1.765*** −0.241*** (0.128) (0.0317) (0.186) (0.0180) (0.196) (0.0158) ESP 0.405*** 0.0882*** −0.0947 −0.0118 0.492*** 0.106*** (0.124) (0.0253) (0.155) (0.0200) (0.135) (0.0309) FRA 0.491*** 0.106*** −0.527*** −0.0742*** 0.287** 0.0601** (0.119) (0.0236) (0.138) (0.0221) (0.13) (0.0284) ISR −0.0576 −0.0133 −0.594*** −0.0856*** −1.371*** −0.205*** (0.117) (0.0273) (0.142) (0.0236) (0.16) (0.0164) JPN −0.656*** −0.159*** −2.254*** −0.443*** −1.985*** −0.256*** (0.111) (0.0274) (0.126) (0.0291) (0.176) (0.0124) KOR 0.311*** 0.0688*** −1.672*** −0.308*** 1.187*** 0.273*** (0.117) (0.0247) (0.135) (0.0313) (0.138) (0.0337) NLD −0.0129 −0.002 96 −0.115 −0.0145 0.532*** 0.116*** (0.116) (0.0269) (0.142) (0.0185) (0.129) (0.03) Constant −5.000*** −0.276 −3.774*** (0.466) (0.168) (0.346) Observations 8605 8605 10 951 10 951 7334 7334 increases with age for energy-efficient light bulbs more likely to invest in RES than younger people and (Mills and Schleich 2012 and 2014). As in Mills and Willis et al (2011) suggest that households with mem- Schleich (2009), age did not seem to be a relevant vari- bers older than 65 years are less likely to adopt solar able for investments in solar panels. Sardianou and technologies compared to the rest of the population. Genoudi (2013) find that middle-aged people are Overall, the impact of age on the probability of 9 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 7. Bayesian Model Averaging Estimates. Logit regressions II. Dependent variables: investments in solar panels, heat pumps, thermal insulation and energy-efficient windows. Solar panels Heat pumps Thermal Insulation Energy-efficient windows Marginal Marginal Marginal Marginal Variables Coefficients effects Coefficients effects Coefficients effects Coefficients effects Age −0.0223*** −0.000 510*** 0.009 30*** 0.002 04*** 0.0149*** 0.003 35*** (0.004 49) (0.000 105) (0.002 21) (0.000 484) (0.002 25) (0.000 506) HHsize 0.213*** 0.0135*** (0.0369) (0.002 34) Education Log_Income 0.136 0.003 11 0.232*** 0.0508*** 0.211*** 0.0474*** (0.126) (0.002 88) (0.0490) (0.0107) (0.0512) (0.0115) NoCope Owner 0.420*** 0.008 84*** 0.687*** 0.141*** 0.612*** 0.131*** (0.155) (0.003 02) (0.0715) (0.0136) (0.0707) (0.0143) House 0.330*** 0.0204*** 0.557*** 0.119*** 0.109 0.0245 (0.11) (0.006 61) (0.0657) (0.0137) (0.0671) (0.015) Tenure −0.122*** −0.0267*** −0.105*** −0.0236*** (0.0246) (0.005 38) (0.0247) (0.005 56) Rural −0.249** −0.0153** 0.0884 0.002 04 0.008 45 0.0019 (0.109) (0.0065) (0.132) (0.003 10) (0.0646) (0.0145) Green_Growther −0.531*** −0.0104*** (0.185) (0.003 14) Altruist −0.986*** −0.0225*** (0.139) (0.003 29) Sceptics 0.131** 0.0289** (0.0578) (0.0128) NGO 0.292*** 0.0638*** 0.244*** 0.0546*** (0.0567) (0.0123) (0.0566) (0.0126) ENV_NGO 0.672*** 0.0536*** 0.694*** 0.0209*** (0.125) (0.0122) (0.170) (0.006 59) Understand_CC −0.099 −0.006 17 (0.106) (0.006 46) Behaviour Index 0.0998*** 0.006 34*** 0.141*** 0.0309*** 0.103*** 0.0232*** (0.0265) (0.001 67) (0.0164) (0.003 58) (0.0162) (0.003 64) KWatt_know Ebill_know −0.0274 −0.001 74 0.0555 0.0125 (0.0941) (0.006) (0.0626) (0.014) Metered 0.113 0.0249 (0.117) (0.0255) Cost_Bias −0.389*** −0.0236*** (0.096) (0.005 54) AUS 1.214*** 0.120*** −2.071*** −0.0242*** 0.776*** 0.184*** −1.829*** −0.290*** (0.24) (0.0337) (0.264) (0.002 40) (0.138) (0.0341) (0.156) (0.0149) CAN −0.567* −0.0296** −1.881*** −0.0234*** 0.111 0.0246 0.347*** 0.0810*** (0.291) (0.0123) (0.243) (0.002 40) (0.133) (0.0300) (0.123) (0.0298) CHE −0.0449 −0.0028 −1.002*** −0.0158*** 0.451*** 0.104*** 0.386*** 0.0907*** (0.273) (0.0168) (0.208) (0.002 52) (0.142) (0.0342) (0.131) (0.0317) CHL −1.327*** −0.0562*** −3.553*** −0.0323*** 0.127 0.0283 −1.203*** −0.221*** (0.312) (0.0084) (0.547) (0.002 70) (0.152) (0.0343) (0.16) (0.0222) ESP −0.0819 −0.005 06 −2.284*** −0.0262*** −0.531*** −0.107*** 0.460*** 0.108*** (0.266) (0.016) (0.274) (0.002 52) (0.148) (0.0269) (0.131) (0.032) FRA 0.0435 0.0028 −1.285*** −0.0189*** 0.459*** 0.106*** 0.715*** 0.171*** (0.259) (0.0169) (0.202) (0.002 38) (0.135) (0.0326) (0.127) (0.0313) ISR 3.441*** 0.577*** −2.121*** −0.0252*** −0.462*** −0.0940*** −1.481*** −0.257*** (0.233) (0.0499) (0.283) (0.002 56) (0.146) (0.0272) (0.15) (0.0179) JPN −0.321 −0.0181 −3.443*** −0.0308*** −0.590*** −0.117*** −1.097*** −0.204*** (0.297) (0.0148) (0.465) (0.002 59) (0.144) (0.0253) (0.138) (0.0199) KOR 0.0451 0.002 91 −1.854*** −0.0232*** 0.523*** 0.122*** 0.415*** 0.0976*** (0.268) (0.0176) (0.261) (0.002 49) (0.137) (0.0333) (0.133) (0.0324) NLD −0.405 −0.0225* −2.454*** −0.0261*** 0.680*** 0.161*** 1.350*** 0.325*** (0.272) (0.0131) (0.304) (0.002 46) (0.139) (0.0343) (0.141) (0.0321) Constant −4.142*** −2.022 −5.454*** −4.536*** (0.296) (1.331) (0.545) (0.563) Observations 6485 6485 7645 7645 6807 6807 7269 7269 Notes: Standard errors are reported in parentheses. Marginal effects at means of dependent variables, superscripts***, ** and* indicate statistical significance at the 1%, 5% and 10% level, respectively. For dummy variables, the marginal effect shows how the predicted probability of observing that a household invests (y = 1) changes as the dummy variables change from 0 to 1. For instance, owners were 7.8 percentage points more likely than renters to own energy- efficient appliances. For continuous variables, the marginal effect measures the instantaneous rate of change. In other words, it measures the change in the predicted probability of observing that a household invests (y = 1) associated with changes in the explanatory variables (X ), when this change is infinitesimally small. For instance, an infinitesimally small increase in the log of income at the sample mean raises the probability to own energy-efficient appliances by 8.3 percentage points. 10 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Figure 1. Predicted probability of investing in energy efficient appliances depending on income. Values for a representative individual. Source: Values as in table 9. investing in clean technologies seems to be technology probability to invest for low-income levels, but this specific and perhaps sometimes driven by age groups. marginal effect decreases and finally levels off for high Family size is positively related to the probability income levels. In the case of energy-efficient appli- to invest in solar panels and light bulbs, while it is not ances, increasing income from 15 000 $ to 45 000 $ included in the preferred model specification for the would lead to an increase of about 10 percentage other technologies. These results are in line with pre- points in the probability to invest, while the same vious studies which also find that the propensity to increase in income would lead to an increase of only 3 adopt solar technologies and light bulbs increases with percentage points in the probability to invest for an family size and children (Mills and Schleich 2009, individual that starts with 60 000 $. The same pattern Mills and Schleich 2012). Mills and Schleich (2010a), emerges for investments in thermal insulation. Those and Mills and Schleich (2012) suggest that a positive results provide clear evidence for financing con- relationship between family size and technology adop- straints. Low-income households are much more tion holds also for energy-efficient appliances. likely to lack both savings to cover the initial invest- We find evidence for credit constraints for some ment costs for clean energy technologies and access to technologies, as investment depends positively on credit. But this barrier is likely to be much less relevant income, except for light bulbs, solar panels and heat for higher-income individuals. This would explain pumps, for which income was not included in the pre- why income increases have a large effect on the prob- ferred model specification or was not a significant ability to invest for lower-income households, but variable. This is in line with previous studies, many of much less so for higher-income households. which find a positive correlation between income and Other socio-economic characteristics were not the probability to invest in energy conservation mea- included in the preferred model. In particular, empiri- sures or renewable energy technologies (Long 1993, cal results from this study have never shown education Mills and Schleich 2010b, Sardianou and Gen- as a key explanatory variable for technology adoption, oudi 2013), while similar to our study Michelsen and in contrast with many studies in the literature (Mills Madlener (2012) did not find any correlation between and Schleich 2009, Di Maria et al 2010, Mills and income and investment in heat pumps. Our findings Schleich 2010a, Michelsen and Madlener 2012, Mills could suggest that public subsidies for solar panels and and Schleich 2012, Sardianou and Genoudi 2013). heat pumps or other policies have helped to overcome Only in a recent study do Mills and Schleich (2014) credit constraints. find that education has no significant impact on light To better understand the extent of credit con- bulb replacement choices. straints, we examined marginal effects across a range There is clear evidence supporting the idea that of income values. The marginal effect of higher renters may have much weaker incentives to invest income on the probability to invest is decreasing, than owners. Owners are more likely to invest than pointing to financing constraints that are particularly renters in energy-efficient appliances, light bulbs, heat relevant for lower-income households. This can be thermostats, heat pumps, thermal insulation and seen for energy-efficient appliances in figure 1, which energy-efficient windows, with a substantially larger shows how the predicted probability to invest evolves magnitude of the effect for relatively immobile invest- with income for a representative individual, whose ments (such as windows and thermal insulation). characteristics are described in more detail in table 9. Nevertheless, renters do invest frequently in more In essence, binary variables take the value that is most mobile technologies with a shorter life cycle, such as frequently observed in the sample, while continuous energy-efficient appliances and light bulbs as shown in variables are evaluated at the sample mean. An table 8. These results confirm the analysis conducted increase in income leads to a big increase in the in OECD (2013). The owner-effect is also well 11 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 8. Share of renters and owners adopting energy efficiency measures and renewables. Energy-efficient Light Heat Solar Thermal Heat Energy-efficient appliances bulbs pumps panels insulation thermostat windows Renters 0.54 0.78 0.03 0.09 0.23 0.25 0.28 Owners 0.66 0.84 0.04 0.12 0.39 0.37 0.42 documented in the literature (Davis 2010, Gillingham Table 9. Characteristics of the representative indivi- dual for energy-efficient appliances. et al 2012). The characteristics of dwellings seem to be rele- Specified characteristics Characteristics at mean vant for technology adoption. The investment prob- Owner = 1 Energy Behaviour Index ability for light bulbs, heat thermostats, thermal NGO = 1 insulation and energy-efficient windows depends KWatt_know = 0 negatively on the time that households have already Metered = 1 spent in their place. That could indicate that house- holds are more likely to invest in energy upgrades when they first move into their home. To our knowl- energy-efficient windows. For solar panels and heat edge, previous studies did not investigate this aspect, pumps households engaged in an environmental focusing more on other characteristics of dwellings, NGO are more likely to invest than others. Social par- such as when the house was built (Mills and ticipation is not only a significant variable for all tech- Schleich 2009, Michelsen and Madlener 2012) or spa- nologies, but the corresponding marginal effects are tial aspects, such as rural or urban area and climate also quite high. For instance, households involved in zone (Michelsen and Madlener 2012). Our results also NGOs are about 8 percentage points more likely to suggest that owning a detached house, which might be invest in energy-efficient appliances than households seen as an indicator of space availability, increases the who are not in a NGO. Work by Olli et al (2001) and probability of investing in light bulbs, heat thermo- Kahn (2007), as well, finds social context to be an stats, thermal insulation and solar panels. For invest- important predictor of environmental behaviour. ment in light bulbs, Di Maria et al (2010) and Mills and Only for solar panels and heat thermostats are Schleich 2010b provide similar results. those households less likely to invest, who attach a Having to pay in line with energy consumption much larger weight to initial investment costs than to and information about this play a role for investment opportunities to reduce the energy bills later on. This in some technologies. Households are more likely to could be indicative of credit constraints or of bounded invest in energy-efficient appliances when they are rationality, whereby consumers use simplified or metered. Households who were able to provide infor- flawed decision making rules that do not involve a full mation about their energy bill or energy consumption comparison between the costs and benefits of invest- are more likely to invest in light bulbs and energy-effi- ments (Yates and Aronson 1983). However, since a cient appliances. For instance, in the case of energy- bias towards initial investment costs is found to influ- efficient appliances, households who report their ence investment decisions only for a few technologies, energy consumption in kilowatt hours are 7 percen- the data do not seem to provide strong evidence in tage points more likely to own these devices than other favour of the idea that bounded rationality con- households. While results do not necessarily imply sistently deters investment in clean energy causality, they do lend some support to the idea that a technologies. lack of information about their own energy consump- An understanding of the causes of climate change tion can limit households’ uptake of energy efficient and attitudes towards the environment do not seem to technologies. play an important role for investment decisions. The There is strong evidence that households who reg- corresponding variables were not included in the pre- ularly perform low-cost energy conservation measures ferred model specification in most cases, but when are also more likely to spend money to conserve they were, results were rather counter-intuitive. As an energy or use renewables. The investment probability example, households who were grouped in the altruist for all technologies, except heat pumps, depends posi- and green growthers clusters seem to be less likely than tively on the energy behaviour index. others to invest in heat pumps. Estimation results suggest that social context is important for investment decisions. Households who are engaged in a NGO are often more likely to invest. Conclusions and policy implications Such social participation correlates positively with technology adoption for energy-efficient appliances, By adopting energy efficiency and renewable energy light bulbs, heat thermostats, thermal insulation and technologies, households can make an important 12 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt contribution to reducing residential energy demand diffuse energy efficiency measures in a market with a and CO emissions. Therefore, understanding the high share of rental housing. determinants of consumers’ investment choices is To promote energy conservation actions and becoming increasingly important. influence individual decision-making, providing The aim of this study is to provide evidence households with feedback on their energy use and a better understanding of their energy practices can be regarding the determinants of investment in energy efficiency and renewables that have been put forward helpful along with energy labels. Recent research in the literature. The data from the OECD Survey on shows that informing households about their energy Household Environmental Behaviour and Attitudes or water consumption compared with that of similar provides a rich basis for such an investigation. households and providing them with conservation Results provide clear evidence supporting the idea tips can lead to important savings (Allcott 2011, Fer- that renters may have much weaker incentives to raro and Price 2013). Those programmes can be used invest than owners. This effect is found for almost all to encourage the adoption of new technologies such as investment goods studied in this paper, with a sub- energy-efficient appliances and, more generally, stantially larger magnitude for relatively immobile encourage households to engage in energy conserva- investments, such as windows and thermal insulation. tion actions and practices. Labels can be also used to Nevertheless, renters show some propensity to invest provide households with reliable information about in lower-cost technologies that are more mobile, such the performance of energy conservation measures or as energy-efficient appliances and light bulbs. renewable energy, encouraging them to conserve Moreover, investment depends positively on energy and invest. income and this effect is larger for lower income levels. This is indicative of credit constraints. Many energy Acknowledgments efficiency and renewable investments have high initial investment costs representing a relevant obstacle, The research leading to these results has received especially for low-income households, who are more funding from the People Programme (Marie Curie likely to be credit-constrained. Actions) of the European Union’s Seventh Frame- Technology adoption is also influenced by house- work Programme (FP7/2007-2013) under REA holds’ attitudes and beliefs, as households who are in grant agreement PIEF-GA-2012-331154—project an environmental group or who are ready to engage in PACE (Property Assessed Clean Energy). The low-cost energy conservation practices are also more authors would like to thank Walid Oueslati, likely to invest in energy efficiency or renewables. Giuseppe Nicoletti, Ysé Serret, Nick Johnstone, These results suggest that targeted policies are Jérôme Silva, Daniel Kammen and various partici- required to address specific barriers for different pants of OECD seminars for their valuable com- groups of consumers. For instance, credit constraints ments and suggestions. The opinions expressed are more relevant for low-income households and lift- herein are those of the authors and do not ing these constraints would likely promote investment necessarily reflect the opinions of the OECD. for this group. Direct subsidies, tax credits or rebates can also be relevant policy instruments to lower the References upfront cost of energy investments. While internalis- ing external costs of emissions by increasing energy Allcott H 2011 Social norms and energy conservation Journal of prices is thought to be a more efficient instrument in Public Economics 95 1082–95 the absence of credit constraints, subsidies to adopt Attari S, DeKay M, Davidson C and de Bruin W 2010 Public perceptions of energy consumption and savings PNAS Early low-emission technologies may be a more effective Edition 107 16054–9 and less costly than higher energy taxation when credit Bulkeley H and Broto V C 2013 Government by experiment? Global constraints are present. cities and the governing of climate change Trans. Inst. Br. The split incentive problem that arises in the rental Geogr. 38 361–75 Davis L W 2010 Evaluating the slow adoption of energy efficient housing market also requires specific policy actions. If investments: are renters less likely to have energy efficient investments in energy-efficient measures were capita- appliances? National Bureau of Economic Research working lised in the purchase and rental prices of the corre- paper 16114 (www.nber.org/papers/w16114.pdf) sponding property, the owner could recover the Di Maria C, Ferreira S and Lazarova E 2010 Shedding light on the light bulb puzzle: the role of attitudes and perceptions in the investment cost. However, this is often not the adoption of energy efficient light bulbs Scottish J. Political case, not least because the effectiveness of energy effi- Econ. 57 48–67 ciency measures is difficult to observe. Energy effi- Donaldson S and Grant-Vallone E 2002 Understanding self-report ciency labels can help in this respect. In addition, bias in organizational behaviour research J. Bus. Psychol. 17 245–60 explicitly allowing owners to increase the rent after Ferraro P J and Price M K 2013 Using nonpecuniary strategies to implementing energy efficiency measures might be influence behavior: evidence from a large-scale field necessary in some countries, where owners are not experiment Rev. Econ. Stat. 95 64–73 allowed to raise the rent as they wish, unless the tenant Gillingham K, Harding M and Rapson D 2012 Split incentives and household energy consumption Energy J. 33 37–62 changes. In Germany, this seems to have helped to 13 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Hassett K A and Metcalf G E 1995 Energy tax credits and residential Morey E, Thacher J and Breffle W 2006 Using angler characteristics conservation investment: evidence from panel data J. Public and attitudinal data to identify environmental preference Econ. 57 201–17 classes: a latent-class model Environ. Res. Econ. 34 IEA 2013 World Energy Outlook (Paris: OECD) (www. 539–65 worldenergyoutlook.org/publications/weo-2013) OECD 2013 Greening household behaviour: overview from the Jaffe A B and Stavins R N 1995 Dynamic incentives of 2011 survey OECD Studies on Environmental Policy and environmental regulations: the effects of alternative policy Household Behaviour (Paris: OECD) (doi:10.1787/ instruments on technology diffusion J. Environ. Econ. 9789264181373-en) Manage. 29 S43–63 Olli E, Grendstad G and Wollebaek D 2001 Correlates of KahnME2007Dogreensdrivehummers orhybrids?environmental environmental behaviour: bringing back social context ideologyasadeterminantof consumer choiceJ.Environ.Econ. Environ. Behav. 33 181–208 Manage. 54 129–45 Poortinga W, Steg L, Vlek C and Wiersma G 2003 Household Kahneman D and Tversky A 1979 Prospect theory: an analysis of preferences for energy-saving measures: a conjoint analysis J. decision under risk Econometrica 47 263–92 Econ. Psychol. 24 49–64 King M F and Bruner G C 2000 Social desirability bias: a neglected Raftery A 1995 Bayesian model averaging selection in social research aspect of validity testing Psychology & Marketing 17 79–103 Soc. Methodol. 25 111–63 Kollmuss A and Agyeman M 2002 Mind the gap: why people act Sardianou E 2007 Estimating energy conservation patterns of Greek environmentally, and what are the barriers to pro- households Energy Policy 35 3778–91 environmental behaviours? Environ. Educ. Res. 8 239–60 Sardianou E and Genoudi P 2013 Which factors affect the Long J 1993 An econometric analysis of residential expenditures on willingness of consumers to adopt renewables energies? energy conservation and renewable energy sources Energy Renew. Energy 57 1–4 Economics 15 232–8 Scott S 1997 Household energy efficiency in Ireland: a replication Mahapatra K and Gustavsson L 2008 An adopter-centric approach study of owner of energy savings items Energy Econ. 19 to analyze the diffusion patterns of innovative residential 187–208 heating systems in Sweden Energy Policy 36 577–90 Shove E 2012 Habits and their creatures The Habits Of Consumption McGuirk P, Bulkeley H and Dowling R 2014 Practices, programs ed A Warde and D Southerton, vol 12 (Helsinki: Collegium) and projects of urban carbon governance: perspectives from pp 100–13 the Australian city Geoforum 52 137–47 Shove E 2014 Putting practise into policy: reconfiguring questions Michelsen C C and Madlener R 2012 Homeowner’s preferences for of consumption and climate change Contemporary Social adopting innovative residential heating system: a discrete Science: J. Acad. Soc. Sci. 9 415–29 choice analysis for Germany Energy Economics 34 1271–83 Simon H A 1959 Theories of decision-making in economics and Mills B and Schleich J 2009 Profits or preferences? assessing the behavioral science Am. Econ. Rev. 49 253–83 adoption of residential solar thermal technologies Energy Sjoberg L and Engelberg E 2005 Lifestyles, and risk perception Policy 37 4145–54 consumer behavior Int. Rev. Sociol.: Revue Internationale de Mills B and Schleich J 2010a What’s driving energy efficient Sociologie 15 327–62 appliances label awareness and purchase propensity? Energy Viklund M 2004 Energy policy options—from the perspective of Policy 38 814–25 public attitudes and risk perceptions Energy Policy 32 Mills B and Schleich J 2010b Why don’t households see the light? 1169–71 Explaining the diffusion of compact fluorescent lamps Willis K, Scarpa R, Gilroy R and Hamza N 2011 Renewable energy Resour. Energy Econ. 32 363–78 adoption in an ageing population: heterogeneity in Mills B and Schleich J 2012 Residential energy-efficient technology preferences for micro-generation technology adoption adoption, energy conservation, knowledge, and attitudes: an Energy Policy 39 6021–9 analysis of european countries Energy Policy 49 616–28 Yates S and Aronson E 1983 A social psychological perspective on Mills B and Schleich J 2014 Household transitions to energy efficient energy conservation in residential buildings Am. Psychol. 38 lighting Energy Economics 46 151–60 435–44 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Environmental Research Letters IOP Publishing

Determinants of households’ investment in energy efficiency and renewables: evidence from the OECD survey on household environmental behaviour and attitudes

Environmental Research Letters , Volume 10 (4): 14 – Apr 1, 2015

Loading next page...
 
/lp/iop-publishing/determinants-of-households-investment-in-energy-efficiency-and-0902urNUrP

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Copyright
Copyright © 2015 IOP Publishing Ltd
eISSN
1748-9326
DOI
10.1088/1748-9326/10/4/044015
Publisher site
See Article on Publisher Site

Abstract

Commons Attribution 3.0 licence. This paper provides novel evidence on the main factors behind consumer choices regarding Any further distribution of investments in energy efficiency and renewable energy technologies using the OECD Survey on this work must maintain Household Environmental Behaviour and Attitudes. The empirical analysis is based on the estimation attribution to the author(s) and the title of of binary logit regression models. Empirical results suggest that households’ propensity to invest in the work, journal citation and DOI. clean energy technologies depends mainly on home ownership, income, social context and household energy conservation practices. Indeed, home owners and high-income households are more likely to invest than renters and low-income households. In addition, environmental attitudes and beliefs, as manifest in energy conservation practices or membership in an environmental non-governmental organisation, also play a relevant role in technology adoption. Introduction economic context, such as available subsidies and energy prices (Poortinga et al 2003, Sardianou 2007, Investments in energy efficiency (EE) and renewable Mahaptra and Gustavsson 2008, Mills and energy sources (RES) are key to reduce greenhouse gas Schleich 2009, Di Maria et al 2010, Willis et al 2011, emissions and thus limit climate change. Global Mills and Schleich 2012, Michelsen and Madl- emissions have continued to grow by 44% from 2000 ener 2012, Sardianou and Genoudi 2013). to 2011 (IEA 2013) and policy action to reverse this Previous studies suggest that households’ socio- trend is urgently needed. In most OECD countries, the economic characteristics play a relevant role in tech- residential sector currently accounts for roughly 30% nology adoption. Generally a positive correlation of final energy consumption and of related CO between income and the probability of investing in emissions (IEA 2013). Further improving the effi- energy technologies is observed (Long 1993, Mills and ciency of energy use as well as boosting the energy Schleich 2010b, Sardianou and Genoudi 2013). It production from renewable sources will be essential to should be noted that this is indicative of credit con- make the necessary progress. By adopting energy straints. In the presence of perfect credit markets poor efficiency measures and renewable energy technolo- households should be able to borrow as long as their gies, households can make an important contribution investments are profitable. Generally, individuals with to reducing greenhouse gas emissions. This paper higher levels of education and those with children are seeks to make a contribution to the technology found to be more likely to adopt energy efficient tech- adoption literature by exploring the determinants of nologies (Mahaptra and Gustavsson 2008, Mills and such investments, which can include households’ Schleich 2009, Mills and Schleich 2012, Michelsen and socio-economic characteristics, characteristics of their Madlener 2012, Sardianou and Genoudi 2013). The dwellings, attitudes regarding environmental pro- impact of age on households’ probability to invest is blems or energy saving measures, households’ knowl- less clear. Some authors find that the propensity to edge about their energy consumption or the cost and adopt energy efficient or renewable technologies performance of energy conservation measures and declines with age (Mahaptra and Gustavsson 2008, renewable technologies, along with policy and Mills and Schleich 2009, Mills and Schleich 2012, © 2015 IOP Publishing Ltd Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Michelsen and Madlener 2012), while other results cost relatively easily, while assessing the total present suggest that middle-aged people are more likely to value of energy savings over the life of an investment is adopt such technologies than younger ones (Mills and a more difficult task given the uncertainty surround- Schleich 2010a, Sardianou and Genoudi 2013). ing energy savings and fluctuations in energy prices A number of studies underline the importance of (Jaffe and Stavins 1995, Hassett and Metcalf 1995). dwelling characteristics behind consumer choices. In This salience effect can lead households to give initial particular, the ownership of the primary residence is costs a higher weight than energy savings. Another an important driver of technology adoption. Evi- phenomenon that might explain the stronger empha- dence on the owner-effect is provided by Davis sis on initial costs than future energy savings often (2010) who shows that renters are significantly less observed in consumer behaviour is termed the ‘status likely to have energy-efficient refrigerators, clothes quo bias’. Kahneman and Tversky (1979) suggest that washers or dishwashers than owners, while Gilling- people normally perceive outcomes as losses and ham et al (2012) demonstrate that owner-occupied gains relative to a reference point, usually the status dwellings are more likely to be insulated than renter- quo. The authors’ empirical results suggest that peo- occupied dwellings. Other studies suggest that dwell- ple exhibit loss aversion in decision making under ings’ characteristics, such as location in a rural area uncertainty, giving much more weight to a possible and/or detachment from other houses may be indica- loss than to an equivalent uncertain gain. In the tors of space availability for investing in particular energy efficiency context, loss aversion can partly energy technologies, while the climate zone can influ- explain why consumers do not take up cost-effective ence the performance of specificenergymeasures investments, as they weight the certain initial costs (Mills and Schleich 2009,DiMaria et al 2010,Michel- (the loss) much more strongly than future uncertain sen and Madlener 2012)and thus people’spropensity benefits, even if these are in principle of an equivalent to invest. value. Technology adoption can also be affected by how Attitudes and beliefs may also play a role as a moti- households collect and process information that is vation to invest in addition to pure monetary benefits necessary to assess whether an investment is profit- and costs of an investment. Indeed, several studies find able. Whether consumers know the costs and benefits that people with strong environmental preferences are of different energy solutions, how much energy they more likely to invest in energy conservation technolo- use in their homes, or what rates of return to expect gies (Olli et al 2001, Kollmuss and Agyeman 2002,Di from energy efficiency measures is likely to affect the Maria et al 2010). Kahn (2007) underlines the impor- adoption of energy-efficient technologies (Mills and tance of living in environmentalist communities as a Schleich 2012). Attari et al (2010) provide evidence on driver for environment-friendly behaviour. He shows households’ misperceptions about energy use or sav- that individuals living in communities with a higher ings. They suggest that there is relatively little knowl- share of green party members are more likely to use edge regarding the effectiveness of different energy public transit, purchase green vehicles (e.g. hybrid), saving measures. Such limited knowledge is likely to and consume less gasoline than people in other com- determine the probability to invest in energy efficiency munities. Also Olli et al (2001) suggest that social con- and renewables. Indeed, when consumers are aware of text is important for environmental behaviour, as potential energy savings, the probability of investing in social participation correlates positively with respon- energy conservation measures increases (Scott 1997). sible environmental behaviour. However, although Di Maria et al (2010) find that the over 50% of their several studies suggest that peoples’ attitudes are pre- respondents are not aware of the potential energy sav- dictors of energy saving behaviour, often this relation- ings of compact fluorescent light bulbs compared to ship is weak, explaining only a small part of household the traditional bulbs. energy choices (Viklund 2004, Sjoberg and Engel- Regarding households’ ability to process informa- berg 2005, Di Maria et al 2010). tion, studies often refer to the concept of bounded Households’ energy use and energy conservation rationality suggesting that customers use simplified actions can also be shaped by habits, routines and decision-making processes, by using only a subset of social practices (Shove 2012). Understanding how the available information for complex decisions social practices occur might be essential to promote (Simon 1959). Research on bounded rationality sug- pro-environmental behaviour. Shove (2014) reports gests that individuals are more likely to take into the example of the ‘Cool Biz’ programme in Japan, account aspects that are easy to perceive than those which has been particularly effective in reducing that are difficult to assess, when they make an invest- energy consumption by modifying employees’ prac- ment decision (Yates and Aronson 1983). Indeed, tices of clothing. The programme establishes that gov- consumers tend to perceive the upfront investment ernment buildings would not be heated or cooled between 20 and 28 degree Celsius and male officers We refer to the economic concept of rationality, where individuals would be encouraged to remove jackets and ties in the are perfectly informed and use all the available information correctly to make optimal decisions. summer, while wearing more in the winter. 2 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Finally, energy prices and a favourable policy con- presents and discusses the empirical results. The final text should also affect household technology adoption. section concludes. In particular, recent studies underline the relevance of urban climate governance, that is the policies and measures undertaken by local governments for energy Data description investment to occur (Bulkeley and Broto 2013). Those policies usually include behaviour change initiatives The survey data were collected through an online aiming at promoting environmental friendly beha- questionnaire, the second of its kind, which was viour; demonstration activities promoting technologi- carried out in early 2011, while the first was launched cal innovations; transition programmes supporting in 2008. The more recent survey, which is the basis for social practices and technological systems to lower the analysis in this paper, collects data from a sample energy consumption and advocacy initiatives attempt- of more than 12 000 respondents, approximately 1000 ing to change public and political approaches to households for each country: Australia (shorthand: improve carbon emission reduction (McGuirk AUS), Canada (shorthand: CAN), Chile (shorthand: et al 2014). CHL), France (shorthand: FRA), Israel (shorthand: This paper provides novel evidence on the main ISR), Japan (shorthand: JPN), Korea (shorthand: factors behind consumer choices regarding the adop- KOR), the Netherlands (shorthand: NLD), Spain tion of energy efficiency and renewable energy tech- (shorthand: ESP), Sweden (shorthand: SWE) and nologies. It is based on the OECD Survey on Switzerland (shorthand: CHE). Household Environmental Behaviour and Attitudes For representativeness, the sample was stratified in comprising household data from 11 OECD countries. each country according to different parameters: age, Results suggest that households’ propensity to invest gender, region and socio-economic groups . Age was in clean energy technologies depends on home owner- stratified using the following groups: 18 to 24, 25 to 34, ship, income, as well as attitudes and beliefs. Indeed, 35 to 44, 45 to 54 and 55 to 69. Gender was approxi- home owners and high-income households are more mately half male and female for all countries. Region likely to invest than renters and low-income house- was stratified and quotas created using three to five holds. Households, who demonstrate environmental regions. For income stratification, households’ after- consciousness, e.g. through membership in a non- tax income quintiles were estimated for each country, governmental organisation (NGOs), in particular then responses from the survey income question were when it is environmental, or through regular energy used to fill the quotas. When quotas were filled, conservations actions, are more likely to invest than respondents with these characteristics were stopped others. from completing the questionnaire. The target The contribution to the literature on households’ respondent was between 18 and 70 years of age and technology adoption behaviour is threefold. First, the had influence on household purchasing decisions and unique dataset underlying this study includes data expenditures. Despite rigorous efforts regarding strati- from 11 OECD countries. To our knowledge, this is fication and quota sampling, it is important to the first cross-country analysis of household invest- acknowledge that there may be some respondent char- ment in energy technologies for countries across the acteristics that were not observed and which correlate OECD. Mills and Schleich (2012) analysed the adop- with internet use. This correlation of unobserved tion of residential energy technologies at the Eur- characteristics could introduce a selection bias in the opean level, while other studies focused on sample. More details on the questionnaire design, technology adoption for single countries. Second, the respondent targeting and quota sampling are provided data allow us to account for a rich set of variables, in OECD (2013), annex B. including respondents’ beliefs, attitudes and beha- viour regarding the environment and their knowledge The OECD ran ‘Call for Tender’ to select a survey service provider about their energy use and spending, in addition to specialised in the implementation of large international web-surveys using online consumer panels in different countries. Global Market more commonly investigated factors, such as house- Insite (GMI) was selected to run the survey and respondents were holds’ socio-economic characteristics, dwelling char- recruited from GMI’s in-country panels. In some countries, GMI acteristics and economic variables. This provides partnered with in-country firms with their own panels in order to further increase panel size. All partners were selected on the basis of novel insights into technology adoption behaviour. quality of panel management. Specifically GMI and its partners Third, this study covers seven different technologies, managed their respondents in line with ESOMAR 26, which is a including energy efficiency measures and renewable standard for transparency and accountability in the use of respon- dent panels for web-based survey research. To limit the risk of energy technologies, revealing differences and simila- recruiting ‘professional respondents’, GMI only permitted panellists rities regarding the determinants of investment in to answer and receive compensation for up to five questionnaires these technologies. per year. Moreover, potential respondents who started the ques- tionnaire were asked whether they met the screening criteria (living The paper is structured as follows. The following in non-institutional settings and influential in household financial section presents the data, while the third section pre- decisions). If they did not meet the criteria, they were screened out sents the econometric model. The fourth section of the sample. 3 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 1. Rates of Technology adoption across countries. Appl Bulb Pump Solar Thrm Heat Windows Country Mean Mean Mean Mean Mean Mean Mean Australia 0.69 0.91 0.03 0.20 0.58 0.15 0.13 Canada 0.67 0.87 0.04 0.04 0.38 0.65 0.51 Chile 0.41 0.95 0.01 0.02 0.31 0.06 0.14 France 0.74 0.86 0.05 0.06 0.45 0.44 0.59 Israel 0.59 0.84 0.03 0.67 0.20 0.11 0.13 Japan 0.48 0.48 0.01 0.04 0.20 0.07 0.19 Korea 0.69 0.63 0.03 0.07 0.38 0.59 0.49 Netherlands 0.61 0.89 0.02 0.04 0.49 0.48 0.73 Spain 0.74 0.91 0.02 0.06 0.21 0.47 0.54 Sweden 0.62 0.87 0.16 0.04 0.29 0.34 0.39 Switzerland 0.62 0.79 0.07 0.05 0.37 0.43 0.49 Total 0.62 0.82 0.04 0.11 0.34 0.33 0.38 The aim of this study is to investigate which factors initial investment cost and easy implementation. However, technology adoption varies significantly might drive household decision making when it comes to the adoption of clean energy technologies. The sur- across countries. Israel is the only country showing a high rate of adoption for solar panels (67%), while on vey data provides a good basis for this, as households average 11% of households have invested in solar tech- were asked whether they installed or bought appli- nologies. In the Netherlands relatively large shares of ances that received a top rating in terms of energy effi- households seem to have invested in thermal insula- ciency between 2001 and 2011. The shorthand for the tion, heat thermostats and energy-efficient windows. corresponding variable used in this paper is ‘Appl’; the Australia, as well, shows a high rate of adoption for variable takes a value of 1 for households who invested thermal insulation (58%), while Canadian households and zero for households, who could have invested, but are particularly likely to invest in heat thermostats decided against it. The same variable is constructed for (65%). In general, Japanese and Chilean households low-energy light bulbs (shorthand: Bulb), energy-effi- invest relatively infrequently in most of the technolo- cient windows (double or triple glazing, shorthand: gies considered in this study, except for energy effi- Windows), thermal insulation of walls or the roof cient appliances and low-energy light bulbs. (shorthand: Thrm), heat thermostats (shorthand: Based on the empirical literature and data avail- Heat), solar panels for electricity or hot water (short- ability, factors that might influence the decision to hand: Solar) or ground source heat pumps (short- invest in energy technology have been grouped in four hand: Pump). The survey also includes data regarding different categories: (1) socio-economic character- wind turbine investments. However given the limited istics of households; (2) the characteristics of their number of investors, namely 158 households, those dwelling; (3) households’ attitudes, knowledge and data were not included in the analysis. To study the behaviour regarding the environment; (4) house- determinants of household investment decisions only holds’ knowledge about their energy spending those households were considered who could in prin- and use. ciple have invested, while those who declared that Socio-economic variables available in the house- their house was already equipped or that, as renters, hold data set include the respondent’s age (Age), gen- they were not allowed to invest were not included in der (Female), household size, the number of years of the analysis. Table 1 shows descriptive statistics for the education after high school (Education), annual net adoption of different technologies. household income (Log_Income), the educational Among the investments considered, low-energy status of the household head, operationalised as a light bulbs were particularly frequently adopted across dummy variable for household heads who are highly countries, with more than 80% of households stating qualified professionals (Prime-earner is high skilled that they had bought such bulbs over the last ten years. worker). There is also a dummy variable that takes a Energy efficient appliances were also relatively fre- value of 1 for households stating that they cannot cope quently adopted, by more than 62% of households, with their current income (NoCope). Descriptive sta- while ground source heat pumps were adopted by only tistics are shown in table 2. a small minority of households, 4% across all coun- Households’ average annual net income is tries. Those numbers suggest that technology adop- approximately 37 868 dollars with considerable differ- tion is more likely for investments with relatively low ences in means across countries. Households living in Chile declared the lowest average annual income Heat thermostat is a device that establishes and maintains a desired (13 585 dollars), while households resident in Switzer- temperature automatically or signals a change in temperature for manual adjustment. land declared the highest level of annual average 4 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 2. Socio-economic characteristics of households. Household Prime-earner is high a a a Age Education Income Size Female NoCope skilled worker Country Mean St Mean St Mean St Mean St Mean Mean Mean Australia 42.20 14.16 3.31 3.47 48 700 27 933 2.90 1.48 0.51 0.39 0.19 Canada 43.59 14.18 3.13 2.94 42 026 26 803 2.51 1.18 0.51 0.37 0.16 Chile 37.41 12.41 4.36 3.04 13 585 10 387 3.84 1.56 0.52 0.44 0.36 France 43.18 14.07 2.64 2.36 38 157 17 697 2.74 1.17 0.51 0.44 0.12 Israel 38.3 13.24 3.95 3.45 26 562 15 329 3.63 1.65 0.55 0.43 0.32 Japan 43.67 13.83 4.90 4.26 48 394 28 702 2.99 1.49 0.49 0.29 0.08 Korea 38.53 11.66 3.26 2.38 27 012 13 892 3.49 1.32 0.50 0.33 0.13 Netherlands 45.18 13.72 4.14 3.16 38 708 16 953 2.63 1.18 0.50 0.24 0.21 Spain 41.72 12.77 3.66 3.07 29 360 16 337 2.99 1.11 0.49 0.37 0.25 Sweden 43.63 14.45 2.39 2.51 41 575 19 181 2.39 1.17 0.48 0.34 0.16 Switzerland 44.21 14.14 2.74 2.66 62 278 29 666 2.67 1.37 0.52 0.36 0.09 Total 42.01 13.77 3.50 3.15 37 868 24 681 2.98 1.42 0.51 0.36 0.19 For dummy variables, standard deviation is not computed. Table 3. Characteristics of dwellings. detached house (54%). Higher rates of ownership are a a a House Tenure Owner Rural observed in Spain (70%), Korea (70%) and the Neth- erlands (68%), while relatively many households live Country Mean Mean Sd Mean Mean in a detached house in Australia (83%), Chile (77%) Australia 0.83 9.36 11.27 0.62 0.20 and the Netherlands (75%). On average, households Canada 0.65 10.72 12.43 0.63 0.27 have lived for approximately 13 years in their primary Chile 0.77 12.95 13.94 0.65 0.14 residence, although average tenure is longer in Japan, France 0.61 12.84 13.81 0.61 0.54 around 18 years. Israel 0.32 15.10 15.62 0.67 0.20 A number of variables reflect respondents’ beliefs, Japan 0.60 18.83 16.70 0.58 0.31 attitudes and behaviours regarding the environment. Korea 0.30 8.63 9.10 0.70 0.07 This includes a dummy variable for households that Netherlands 0.75 15.99 14.81 0.68 0.53 Spain 0.26 13.76 12.97 0.80 0.38 participate in a non-governmental organisation Sweden 0.47 10.74 12.34 0.60 0.47 (NGO) and another one for those that are specifically Switzerland 0.36 11.70 12.05 0.38 0.61 in an environmental NGO (Env NGO). There is a Total 0.54 12.86 13.68 0.63 0.34 dummy variable for people who rated the environ- ment as the most pressing concern (Env_top_cncrn) For dummy variables, standard deviation is not computed. and another one for those who instead rated the econ- omy as the most pressing concern (Eco_top_cncrn). income (62 278 dollars). 36% of households stated Another dummy variable is used for those respon- that their salary was not enough to cover their needs, dents who were able to identify the causes of climate and difficulty to cope with income is particularly an change correctly (Understand_CC). issue in Chile (44%), France (44%) and Israel (43%). Respondents were asked questions regarding their It is quite surprising that Chile and Israel are the two willingness to make sacrifices to protect the environ- main countries with a higher percentage of profes- ment, their assessment of the need to do so and the sionals as household heads, 36% and 32%, respec- role of technology in solving environmental problems. tively, while France showed one of the lowest shares Depending on their answers to those questions house- (12%). Those data could partly result from French holds were grouped in three clusters : i) the envir- respondents’ difficulty to classify their occupation onmentally motivated, who are willing to make according to the categories they were given, as they sacrifices in their lifestyle to solve environmental pro- particularly frequently classified their occupation as blems (Altruists), ii) environmental sceptics who are ‘other’. The average length of education after high not willing to make much effort to solve environ- school is 3.5 years, suggesting that a number of mental problems, which they believe are often exag- respondents went to university. gerated (Sceptics), and iii) a group of technological The survey includes some characteristics of dwell- optimists who believe that environmental problems ings, such as home-ownership versus rental (Owner), dwelling type (House), years lived in the primary resi- To uncover these attitudinal profiles the latent class method (LCA) dence (Tenure) and whether households live in a rural is used. LCA is a statistical method for identifying unmeasured class membership among subjects using categorical and/or continuous area (Rural). Table 3 lists the variables used. observed variables. A description and demonstration of LCA in the The majority of respondents (63%) own their resi- context of environmental attitudes can be found in Morey dence and more than half of households live in a et al (2006). 5 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 4. Respondents’ beliefs, attitudes and behaviours regarding the environment. Env Top Eco Top Understand Cost a a a a a b Green Growthers Altruist Sceptics NGO Env NGO Concern Concern CC Bias Country Mean Sd Mean Sd Mean Sd Mean Mean Mean Mean Mean Mean Australia 0.10 0.30 0.42 0.49 0.44 0.50 0.52 0.09 0.18 0.32 0.27 0.35 Canada 0.14 0.35 0.46 0.50 0.37 0.48 0.53 0.10 0.13 0.36 0.30 0.31 Chile 0.26 0.44 0.54 0.50 0.19 0.39 0.69 0.17 0.19 0.27 0.29 0.45 France 0.10 0.30 0.56 0.49 0.33 0.47 0.44 0.08 0.11 0.44 0.24 0.25 Israel 0.12 0.32 0.64 0.48 0.23 0.42 0.50 0.12 0.09 0.21 0.15 0.39 Japan 0.19 0.39 0.31 0.46 0.46 0.50 0.32 0.03 0.17 0.49 0.26 0.36 Korea 0.25 0.44 0.38 0.48 0.36 0.48 0.42 0.05 0.28 0.37 0.21 0.33 Netherlands 0.24 0.42 0.29 0.45 0.43 0.49 0.52 0.12 0.11 0.27 0.32 0.32 Spain 0.22 0.42 0.37 0.48 0.38 0.49 0.57 0.10 0.04 0.62 0.21 0.41 Sweden 0.09 0.29 0.55 0.49 0.34 0.479 0.56 0.11 0.25 0.25 0.44 0.44 Switzerland 0.15 0.36 0.48 0.50 0.36 0.48 0.63 0.20 0.22 0.26 0.33 0.40 Total 0.17 0.38 0.45 0.50 0.35 0.48 0.52 0.11 0.16 0.35 0.27 0.36 For dummy variables, standard deviation is not computed. This variable considers that the rating given to initial investment cost exceeds the rating for future energy prices by three points on the scale. are real and technological innovations are key to sol- with some country variation. This share is 44% of ving them (Green Growthers) (OECD 2013). Respon- households in Sweden and only around 15% in Israel. dents were grouped according to their agreement on At the same time, Swedish households are more likely seven statements: (1) policies introduced by govern- to make sacrifices in their lifestyle to solve environ- ment to address environmental issues should not cost mental problems. On the other hand, Dutch house- me extra money, (2) I am willing to make compro- holds are the less likely to sacrifice their lifestyle for the mises in my current lifestyle for the benefit of the environment, although 32% of them are aware about environment, (3) protecting the environment is a the causes of climate change. Quite surprisingly, Israel shows the lowest level of awareness regarding the cau- means of stimulating economic growth, (4) Environ- mental issues will be resolved in any case through tech- ses of climate change (15%), but at the same time this nological progress, (5) environmental impacts are is the country with the highest percentage of respon- frequently overstated, (6) I am not willing to do any- dents who are environmental ‘Altruist’ (64%). thing about the environment if others do not do the A variable is constructed to capture a bias, in that a same, (7) environmental issues should be dealt with much larger weight is given to initial investment costs than to higher energy prices, although the impact on primarily by future generations. Table 4 summarises households’ propensity to invest should be the same the variables related to social context and environ- mental behaviour. for investment cost reductions and energy price The percentage of respondents who believe that increases that have an equivalent present value. environmental problems are real and express a will- Households were asked to rate different reasons that ingness to make compromises in their lifestyle to solve would induce them to invest in energy efficiency or them is 45%, although with some country variation, as change their behaviour to save more energy on a scale of 0 to 10. A variable named ‘Cost bias’ identifies 64% of respondents are ‘Altruists’ in Israel and around 55% in France and Sweden. On the other hand, on households giving a significantly higher rating to lower average across countries 35% of respondents are scep- initial investment costs than to higher energy prices. When the rating given to initial investment cost tical about environmental problems. Japan showed the highest level of scepticism (46%), while Chile exceeds the rating for future energy prices by three showed the lowest level (19%). This mirrors the share points on the scale, respondents are considered as hav- of environmental ‘Altruists’ in those two countries, ing a bias towards initial investment costs and the dummy variable takes a value of 1. As a sensitivity test, 31% in Japan and 54% in Chile. In most countries, more than 50% of respondents the variable is also constructed considering a differ- are engaged in some non-governmental organisation ence of 4 points on the scale. On average, 36% of respondents give a rating to initial investment costs (NGO). Only in Japan is this share much lower, just above 30%. On average across countries around 10% that is 3 points higher than the rating for energy prices, of respondents are engaged in an environmental 30% give a rating to initial investment costs that is 4 points higher. The highest rate is observed for Chile, NGO, but both in Japan and Korea this share is much lower. which is also the country with the lowest average Less than one third of respondents (27%) seemed annual income. This could suggest that financial con- to understand the causes of climate change, although straints might partly explain why households give a 6 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 5. Households’ knowledge about their energy spending and use. a a a Metered Ebill_known KWatt_known Behaviour Index Country Mean Mean Mean Mean Sd Australia 0.91 0.66 0.14 7.91 1.66 Canada 0.81 0.53 0.12 7.10 1.77 Chile 0.91 0.71 0.18 8.32 1.63 France 0.94 0.66 0.15 7.95 1.59 Israel 0.89 0.65 0.13 7.64 1.69 Japan 0.97 0.58 0.21 7.08 1.86 Korea 0.96 0.12 0.26 7.80 1.70 Netherlands 0.91 0.38 0.26 7.06 1.75 Spain 0.93 0.63 0.13 8.39 1.45 Sweden 0.90 0.56 0.37 5.55 1.84 Switzerland 0.90 0.63 0.15 6.82 1.77 Total 0.91 0.55 0.19 7.43 1.86 For dummy variables, standard deviation is not computed. stronger weight to investment costs than to future When interpreting these results, it is important to energy prices. However, a high percentage of biased keep in mind that response bias might occur. The wording of questions, the response scale (especially for consumers is also observed in Switzerland, the country attitudinal questions), context and data collection with the highest level of average annual income. techniques can all affect the way responses are pro- The data also capture households’ knowledge vided. In particular, even the slightest suggestion in the about their energy spending and use. A large majority way a question regarding opinions and attitudes is for- of respondents, 91% on average across countries, sta- mulated can potentially lead the respondent toward a ted that their energy consumption is metered. It is not particular answer. For instance, some results concern- very common for households to be informed about ing households’ rationality could reflect the respon- their energy bills and use, though. Respondents were dents’ difficulty to interpret correctly the questions asked to get hold of their energy bills before answering themselves. In addition, self-report bias can occur. the survey, but only about 55% were able to provide Generally, survey participants have the tendency to information about their energy spending on average respond in a way that makes them look as good as pos- across countries. Even fewer households were able to sible, or socially desirable, and this could have a bear- provide information about their energy consumption, ing on the survey findings (King and Bruner 2000, less than 19% on average across countries. The data Donaldson and Grant-Vallone 2002). Another source show an unusual result for Korean households, who of bias can be strategic misrepresentation. When seem better informed about their energy consumption respondents expect a possible connection between in volumes than about their energy spending. their response and some economic outcome in which Regarding energy use, the ‘behaviour index’ vari- they have an interest, they may have strategic incen- able captures whether respondents perform certain tives to misrepresent information. For instance, when energy conservation actions regularly, such as turning households are asked whether high energy prices off the lights when leaving the room, cutting down on would induce them to invest in energy efficiency, they heating/air conditioning to limit energy consumption, may understate their reaction to energy prices if they believe that their response could lead to an increase in running full loads when using the washing machines, future energy prices, perhaps because they think that washing clothes using cold rather than warm/hot survey results might induce the government to raise water, switching off the standby mode of appliances energy taxes. and air dry laundry rather than using a clothes dryer. The behaviour index ranges from 0 to 10, where higher values indicate that households perform several of Method these actions regularly. The data suggest that house- holds perform quite regularly energy conservation Households’ investment in energy efficiency and actions and on average across countries the behaviour renewables is investigated within a discrete choice modelling framework. For each investment good i index takes values around 7. Lower values of the index studied in this paper, households’ investment is are observed in Sweden (5.55) and Switzerland (6.82), modelled as: while higher values are observed in Chile (8.32) and Spain (8.39). Table 5 summarises the variables related yx=+βϵ (1) ii to household’s knowledge about their energy spending and use. 7 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt where y is a latent variable that captures households’ evaluate marginal effects this way and we follow this preference for technology i, namely the difference practice in most cases. Yet, as marginal effects may be between the marginal benefit and the marginal cost of very different at data points that are different from the adopting this good. X is a vector of explanatory i sample mean, it can be useful to examine marginal variables (e.g. socio-economic characteristics, dwell- effects across a range of values for some explanatory ings’ characteristics, households’ attitudes, knowledge variables, which we explore for income. In this exer- and behaviour, and household’s energy use), β is the cise dummy variables take the value that is most fre- parameter vector to be estimated and ϵ is the error quently observed in the sample, while only continuous term. While preferences cannot be observed directly, variables, except income, are evaluated at the sample the decision to adopt technology i can be observed and mean. The logit model is estimated with country-level it is modelled in line with the following decision rule: fixed effects that capture differences in policy and eco- nomic context, income levels and other country-spe- yi=<00 fy i i cific circumstances. Sweden is the base country. yi=⩾10 fy (2) The Bayesian model averaging (BMA) method is used to determine the best model specification. In That is, a household invests in good i y = 1 if ( ) absence of a theoretical model, BMA offers a systema- the marginal benefit of adopting this good is larger tic method for analysing specification uncertainty and than or equal to the marginal cost, otherwise it does checking the robustness of results to alternative model not invest(0 y = ). The probability of households’ specifications (Raftery 1995). For each tested explana- investing in good i is modelled as follows: tory variable, BMA provides the probability that this exp x β () i variable is included in the true model. This is calcu- Py== 1 x () lated on the basis of weights assigned to each tested 1e + xp x β () i models. The BMA method selects the ‘best’ model (the =Λβ x (3) () one with highest posterior probability) based on all possible combinations of the explanatory variables. In where Λ denotes the logistic cumulative distribution this paper, BMA also helps to deal with collinearity function. issues in the canonical regressions, which include all Given the non-linearity of the logit model, mar- explanatory variables. In particular, collinearity occurs ginal effects have to be calculated from the underlying with attitudinal variables, such as altruists, green estimates. For continuous variables, the marginal growthers and sceptics, while in the model selected effect measures the change in the predicted probability with the BMA method those variables are never inclu- of observing that a household invests (y = 1) associated ded all together. with changes in the explanatory variables (X ) that are infinitesimally small. For dummy variables, the mar- ginal effect measures how the predicted probability of Results and discussion observing that a household invests (y = 1) changes as the dummy variables change from 0 to 1. It should be This section discusses results from the preferred model noted that marginal effects do not have a clear inter- selected with the BMA method (tables 6 and 7). Some pretation for ordinal variables, such as the behaviour variables which have been never included in the index, as differences between different levels of this preferred model are not reported in the tables. These variable are not meaningful. Only the sign of the mar- variables include the occupation of the household ginal effect can be interpreted in this case. head and an index variable for those respondents who In this study, marginal effects are evaluated at the rated the environment or economy as the most sample means of the independent variables. In parti- pressing concern. Results from the second and third cular, marginal effects for continuous variables are best model, along with canonical regressions that computed as follows: contain all available explanatory variables, can be made available upon request. ∂= Pryx 1 () i Results suggest that socio-economic character- ⎡ ⎤ =  Λβxx1( − Λ β  β4) ()ii( ) ⎣ ⎦ istics of households partly explain investment in ∂x energy efficiency and renewables. The age of the respondent appears to be a relevant variable for most For dummy variables, the marginal effect is the difference between the predicted probability to invest of the technologies analysed. Investments in light bulbs, heat thermostats, thermal insulation and with the dummy taking a value of 1 and of zero, when all other independent variables, including dummy energy-efficient windows depend positively on age, while the probability to choose heat pumps decreases variables, are evaluated at the sample mean. While this is not very intuitive, considering that dummy variables with age. Earlier studies confirm that the probability of investing declines with age for innovative heating sys- for actual observations take either the value 0 or 1, but never the sample mean—as an example, nobody can tems, such as heat pumps, (Mahaptra and Gus- possibly be 52% female—it is common practice to tavsson 2008, Michelsen and Madlener 2012), while it 8 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 6. Bayesian Model Averaging Estimates. Logit regressions I. Dependent variables: investments in energy-efficient appliances, light bulbs and heat thermostat. Energy-efficient appliances Light bulbs Heat thermostat Variables Coefficients Marginal effects Coefficients Marginal effects Coefficients Marginal effects Age 0.0140*** 0.001 71*** 0.0115*** 0.002 29*** (0.002 26) (0.000 274) (0.002 34) (0.000 465) HHsize 0.135*** 0.0165*** (0.0249) (0.003 02) Education Log_Income 0.362*** 0.0833*** 0.0154*** 0.003 06*** (0.0425) (0.009 78) (0.0027) (0.000 538) NoCope Owner 0.336*** 0.0783*** 0.151** 0.0186** 0.372*** 0.0714*** (0.0518) (0.0122) (0.0628) (0.007 89) (0.073) (0.0134) House 0.295*** 0.0363*** 0.167** 0.0330** (0.0656) (0.008 14) (0.0694) (0.0136) Tenure −0.0502** −0.006 11** −0.0908*** −0.0181*** (0.0247) (0.003 01) (0.0258) (0.005 13) Rural 0.0206 0.002 50 0.150** 0.0301** (0.0651) (0.007 89) (0.0654) (0.0133) Green_Growther Altruist Sceptics −0.200*** −0.0248*** (0.0566) (0.007 19) NGO 0.345*** 0.0797*** 0.416*** 0.0511*** 0.270*** 0.0535*** (0.0485) (0.0112) (0.0563) (0.006 96) (0.0584) (0.0115) ENV_NGO Understand_CC Behaviour Index 0.140*** 0.0322*** 0.161*** 0.0196*** 0.0916*** 0.0182*** (0.0139) (0.003 19) (0.0154) (0.001 86) (0.0167) (0.003 32) KWatt_know 0.316*** 0.0708*** −0.007 62 −0.00 151 (0.0604) (0.0131) (0.0743) (0.0148) Ebill_know 0.189*** 0.0232*** −0.0938 −0.0188 (0.0592) (0.007 34) (0.0655) (0.0132) Metered 0.354*** 0.0845*** (0.0939) (0.023) Cost_Bias −0.125** −0.0247** (0.0599) (0.0117) AUS 0.0831 0.0189 −0.0286 −0.003 51 −1.214*** −0.185*** (0.121) (0.0273) (0.165) (0.0205) (0.153) (0.0165) CAN 0.158 0.0357 −0.283** −0.0374* 1.157*** 0.266*** (0.117) (0.0259) (0.141) (0.0202) (0.129) (0.0316) CHE −0.109 −0.0255 −0.755*** −0.114*** 0.263* 0.0550* (0.12) (0.0283) (0.132) (0.0237) (0.136) (0.0296) CHL −0.690*** −0.167*** 0.365** 0.0400** −1.765*** −0.241*** (0.128) (0.0317) (0.186) (0.0180) (0.196) (0.0158) ESP 0.405*** 0.0882*** −0.0947 −0.0118 0.492*** 0.106*** (0.124) (0.0253) (0.155) (0.0200) (0.135) (0.0309) FRA 0.491*** 0.106*** −0.527*** −0.0742*** 0.287** 0.0601** (0.119) (0.0236) (0.138) (0.0221) (0.13) (0.0284) ISR −0.0576 −0.0133 −0.594*** −0.0856*** −1.371*** −0.205*** (0.117) (0.0273) (0.142) (0.0236) (0.16) (0.0164) JPN −0.656*** −0.159*** −2.254*** −0.443*** −1.985*** −0.256*** (0.111) (0.0274) (0.126) (0.0291) (0.176) (0.0124) KOR 0.311*** 0.0688*** −1.672*** −0.308*** 1.187*** 0.273*** (0.117) (0.0247) (0.135) (0.0313) (0.138) (0.0337) NLD −0.0129 −0.002 96 −0.115 −0.0145 0.532*** 0.116*** (0.116) (0.0269) (0.142) (0.0185) (0.129) (0.03) Constant −5.000*** −0.276 −3.774*** (0.466) (0.168) (0.346) Observations 8605 8605 10 951 10 951 7334 7334 increases with age for energy-efficient light bulbs more likely to invest in RES than younger people and (Mills and Schleich 2012 and 2014). As in Mills and Willis et al (2011) suggest that households with mem- Schleich (2009), age did not seem to be a relevant vari- bers older than 65 years are less likely to adopt solar able for investments in solar panels. Sardianou and technologies compared to the rest of the population. Genoudi (2013) find that middle-aged people are Overall, the impact of age on the probability of 9 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 7. Bayesian Model Averaging Estimates. Logit regressions II. Dependent variables: investments in solar panels, heat pumps, thermal insulation and energy-efficient windows. Solar panels Heat pumps Thermal Insulation Energy-efficient windows Marginal Marginal Marginal Marginal Variables Coefficients effects Coefficients effects Coefficients effects Coefficients effects Age −0.0223*** −0.000 510*** 0.009 30*** 0.002 04*** 0.0149*** 0.003 35*** (0.004 49) (0.000 105) (0.002 21) (0.000 484) (0.002 25) (0.000 506) HHsize 0.213*** 0.0135*** (0.0369) (0.002 34) Education Log_Income 0.136 0.003 11 0.232*** 0.0508*** 0.211*** 0.0474*** (0.126) (0.002 88) (0.0490) (0.0107) (0.0512) (0.0115) NoCope Owner 0.420*** 0.008 84*** 0.687*** 0.141*** 0.612*** 0.131*** (0.155) (0.003 02) (0.0715) (0.0136) (0.0707) (0.0143) House 0.330*** 0.0204*** 0.557*** 0.119*** 0.109 0.0245 (0.11) (0.006 61) (0.0657) (0.0137) (0.0671) (0.015) Tenure −0.122*** −0.0267*** −0.105*** −0.0236*** (0.0246) (0.005 38) (0.0247) (0.005 56) Rural −0.249** −0.0153** 0.0884 0.002 04 0.008 45 0.0019 (0.109) (0.0065) (0.132) (0.003 10) (0.0646) (0.0145) Green_Growther −0.531*** −0.0104*** (0.185) (0.003 14) Altruist −0.986*** −0.0225*** (0.139) (0.003 29) Sceptics 0.131** 0.0289** (0.0578) (0.0128) NGO 0.292*** 0.0638*** 0.244*** 0.0546*** (0.0567) (0.0123) (0.0566) (0.0126) ENV_NGO 0.672*** 0.0536*** 0.694*** 0.0209*** (0.125) (0.0122) (0.170) (0.006 59) Understand_CC −0.099 −0.006 17 (0.106) (0.006 46) Behaviour Index 0.0998*** 0.006 34*** 0.141*** 0.0309*** 0.103*** 0.0232*** (0.0265) (0.001 67) (0.0164) (0.003 58) (0.0162) (0.003 64) KWatt_know Ebill_know −0.0274 −0.001 74 0.0555 0.0125 (0.0941) (0.006) (0.0626) (0.014) Metered 0.113 0.0249 (0.117) (0.0255) Cost_Bias −0.389*** −0.0236*** (0.096) (0.005 54) AUS 1.214*** 0.120*** −2.071*** −0.0242*** 0.776*** 0.184*** −1.829*** −0.290*** (0.24) (0.0337) (0.264) (0.002 40) (0.138) (0.0341) (0.156) (0.0149) CAN −0.567* −0.0296** −1.881*** −0.0234*** 0.111 0.0246 0.347*** 0.0810*** (0.291) (0.0123) (0.243) (0.002 40) (0.133) (0.0300) (0.123) (0.0298) CHE −0.0449 −0.0028 −1.002*** −0.0158*** 0.451*** 0.104*** 0.386*** 0.0907*** (0.273) (0.0168) (0.208) (0.002 52) (0.142) (0.0342) (0.131) (0.0317) CHL −1.327*** −0.0562*** −3.553*** −0.0323*** 0.127 0.0283 −1.203*** −0.221*** (0.312) (0.0084) (0.547) (0.002 70) (0.152) (0.0343) (0.16) (0.0222) ESP −0.0819 −0.005 06 −2.284*** −0.0262*** −0.531*** −0.107*** 0.460*** 0.108*** (0.266) (0.016) (0.274) (0.002 52) (0.148) (0.0269) (0.131) (0.032) FRA 0.0435 0.0028 −1.285*** −0.0189*** 0.459*** 0.106*** 0.715*** 0.171*** (0.259) (0.0169) (0.202) (0.002 38) (0.135) (0.0326) (0.127) (0.0313) ISR 3.441*** 0.577*** −2.121*** −0.0252*** −0.462*** −0.0940*** −1.481*** −0.257*** (0.233) (0.0499) (0.283) (0.002 56) (0.146) (0.0272) (0.15) (0.0179) JPN −0.321 −0.0181 −3.443*** −0.0308*** −0.590*** −0.117*** −1.097*** −0.204*** (0.297) (0.0148) (0.465) (0.002 59) (0.144) (0.0253) (0.138) (0.0199) KOR 0.0451 0.002 91 −1.854*** −0.0232*** 0.523*** 0.122*** 0.415*** 0.0976*** (0.268) (0.0176) (0.261) (0.002 49) (0.137) (0.0333) (0.133) (0.0324) NLD −0.405 −0.0225* −2.454*** −0.0261*** 0.680*** 0.161*** 1.350*** 0.325*** (0.272) (0.0131) (0.304) (0.002 46) (0.139) (0.0343) (0.141) (0.0321) Constant −4.142*** −2.022 −5.454*** −4.536*** (0.296) (1.331) (0.545) (0.563) Observations 6485 6485 7645 7645 6807 6807 7269 7269 Notes: Standard errors are reported in parentheses. Marginal effects at means of dependent variables, superscripts***, ** and* indicate statistical significance at the 1%, 5% and 10% level, respectively. For dummy variables, the marginal effect shows how the predicted probability of observing that a household invests (y = 1) changes as the dummy variables change from 0 to 1. For instance, owners were 7.8 percentage points more likely than renters to own energy- efficient appliances. For continuous variables, the marginal effect measures the instantaneous rate of change. In other words, it measures the change in the predicted probability of observing that a household invests (y = 1) associated with changes in the explanatory variables (X ), when this change is infinitesimally small. For instance, an infinitesimally small increase in the log of income at the sample mean raises the probability to own energy-efficient appliances by 8.3 percentage points. 10 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Figure 1. Predicted probability of investing in energy efficient appliances depending on income. Values for a representative individual. Source: Values as in table 9. investing in clean technologies seems to be technology probability to invest for low-income levels, but this specific and perhaps sometimes driven by age groups. marginal effect decreases and finally levels off for high Family size is positively related to the probability income levels. In the case of energy-efficient appli- to invest in solar panels and light bulbs, while it is not ances, increasing income from 15 000 $ to 45 000 $ included in the preferred model specification for the would lead to an increase of about 10 percentage other technologies. These results are in line with pre- points in the probability to invest, while the same vious studies which also find that the propensity to increase in income would lead to an increase of only 3 adopt solar technologies and light bulbs increases with percentage points in the probability to invest for an family size and children (Mills and Schleich 2009, individual that starts with 60 000 $. The same pattern Mills and Schleich 2012). Mills and Schleich (2010a), emerges for investments in thermal insulation. Those and Mills and Schleich (2012) suggest that a positive results provide clear evidence for financing con- relationship between family size and technology adop- straints. Low-income households are much more tion holds also for energy-efficient appliances. likely to lack both savings to cover the initial invest- We find evidence for credit constraints for some ment costs for clean energy technologies and access to technologies, as investment depends positively on credit. But this barrier is likely to be much less relevant income, except for light bulbs, solar panels and heat for higher-income individuals. This would explain pumps, for which income was not included in the pre- why income increases have a large effect on the prob- ferred model specification or was not a significant ability to invest for lower-income households, but variable. This is in line with previous studies, many of much less so for higher-income households. which find a positive correlation between income and Other socio-economic characteristics were not the probability to invest in energy conservation mea- included in the preferred model. In particular, empiri- sures or renewable energy technologies (Long 1993, cal results from this study have never shown education Mills and Schleich 2010b, Sardianou and Gen- as a key explanatory variable for technology adoption, oudi 2013), while similar to our study Michelsen and in contrast with many studies in the literature (Mills Madlener (2012) did not find any correlation between and Schleich 2009, Di Maria et al 2010, Mills and income and investment in heat pumps. Our findings Schleich 2010a, Michelsen and Madlener 2012, Mills could suggest that public subsidies for solar panels and and Schleich 2012, Sardianou and Genoudi 2013). heat pumps or other policies have helped to overcome Only in a recent study do Mills and Schleich (2014) credit constraints. find that education has no significant impact on light To better understand the extent of credit con- bulb replacement choices. straints, we examined marginal effects across a range There is clear evidence supporting the idea that of income values. The marginal effect of higher renters may have much weaker incentives to invest income on the probability to invest is decreasing, than owners. Owners are more likely to invest than pointing to financing constraints that are particularly renters in energy-efficient appliances, light bulbs, heat relevant for lower-income households. This can be thermostats, heat pumps, thermal insulation and seen for energy-efficient appliances in figure 1, which energy-efficient windows, with a substantially larger shows how the predicted probability to invest evolves magnitude of the effect for relatively immobile invest- with income for a representative individual, whose ments (such as windows and thermal insulation). characteristics are described in more detail in table 9. Nevertheless, renters do invest frequently in more In essence, binary variables take the value that is most mobile technologies with a shorter life cycle, such as frequently observed in the sample, while continuous energy-efficient appliances and light bulbs as shown in variables are evaluated at the sample mean. An table 8. These results confirm the analysis conducted increase in income leads to a big increase in the in OECD (2013). The owner-effect is also well 11 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Table 8. Share of renters and owners adopting energy efficiency measures and renewables. Energy-efficient Light Heat Solar Thermal Heat Energy-efficient appliances bulbs pumps panels insulation thermostat windows Renters 0.54 0.78 0.03 0.09 0.23 0.25 0.28 Owners 0.66 0.84 0.04 0.12 0.39 0.37 0.42 documented in the literature (Davis 2010, Gillingham Table 9. Characteristics of the representative indivi- dual for energy-efficient appliances. et al 2012). The characteristics of dwellings seem to be rele- Specified characteristics Characteristics at mean vant for technology adoption. The investment prob- Owner = 1 Energy Behaviour Index ability for light bulbs, heat thermostats, thermal NGO = 1 insulation and energy-efficient windows depends KWatt_know = 0 negatively on the time that households have already Metered = 1 spent in their place. That could indicate that house- holds are more likely to invest in energy upgrades when they first move into their home. To our knowl- energy-efficient windows. For solar panels and heat edge, previous studies did not investigate this aspect, pumps households engaged in an environmental focusing more on other characteristics of dwellings, NGO are more likely to invest than others. Social par- such as when the house was built (Mills and ticipation is not only a significant variable for all tech- Schleich 2009, Michelsen and Madlener 2012) or spa- nologies, but the corresponding marginal effects are tial aspects, such as rural or urban area and climate also quite high. For instance, households involved in zone (Michelsen and Madlener 2012). Our results also NGOs are about 8 percentage points more likely to suggest that owning a detached house, which might be invest in energy-efficient appliances than households seen as an indicator of space availability, increases the who are not in a NGO. Work by Olli et al (2001) and probability of investing in light bulbs, heat thermo- Kahn (2007), as well, finds social context to be an stats, thermal insulation and solar panels. For invest- important predictor of environmental behaviour. ment in light bulbs, Di Maria et al (2010) and Mills and Only for solar panels and heat thermostats are Schleich 2010b provide similar results. those households less likely to invest, who attach a Having to pay in line with energy consumption much larger weight to initial investment costs than to and information about this play a role for investment opportunities to reduce the energy bills later on. This in some technologies. Households are more likely to could be indicative of credit constraints or of bounded invest in energy-efficient appliances when they are rationality, whereby consumers use simplified or metered. Households who were able to provide infor- flawed decision making rules that do not involve a full mation about their energy bill or energy consumption comparison between the costs and benefits of invest- are more likely to invest in light bulbs and energy-effi- ments (Yates and Aronson 1983). However, since a cient appliances. For instance, in the case of energy- bias towards initial investment costs is found to influ- efficient appliances, households who report their ence investment decisions only for a few technologies, energy consumption in kilowatt hours are 7 percen- the data do not seem to provide strong evidence in tage points more likely to own these devices than other favour of the idea that bounded rationality con- households. While results do not necessarily imply sistently deters investment in clean energy causality, they do lend some support to the idea that a technologies. lack of information about their own energy consump- An understanding of the causes of climate change tion can limit households’ uptake of energy efficient and attitudes towards the environment do not seem to technologies. play an important role for investment decisions. The There is strong evidence that households who reg- corresponding variables were not included in the pre- ularly perform low-cost energy conservation measures ferred model specification in most cases, but when are also more likely to spend money to conserve they were, results were rather counter-intuitive. As an energy or use renewables. The investment probability example, households who were grouped in the altruist for all technologies, except heat pumps, depends posi- and green growthers clusters seem to be less likely than tively on the energy behaviour index. others to invest in heat pumps. Estimation results suggest that social context is important for investment decisions. Households who are engaged in a NGO are often more likely to invest. Conclusions and policy implications Such social participation correlates positively with technology adoption for energy-efficient appliances, By adopting energy efficiency and renewable energy light bulbs, heat thermostats, thermal insulation and technologies, households can make an important 12 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt contribution to reducing residential energy demand diffuse energy efficiency measures in a market with a and CO emissions. Therefore, understanding the high share of rental housing. determinants of consumers’ investment choices is To promote energy conservation actions and becoming increasingly important. influence individual decision-making, providing The aim of this study is to provide evidence households with feedback on their energy use and a better understanding of their energy practices can be regarding the determinants of investment in energy efficiency and renewables that have been put forward helpful along with energy labels. Recent research in the literature. The data from the OECD Survey on shows that informing households about their energy Household Environmental Behaviour and Attitudes or water consumption compared with that of similar provides a rich basis for such an investigation. households and providing them with conservation Results provide clear evidence supporting the idea tips can lead to important savings (Allcott 2011, Fer- that renters may have much weaker incentives to raro and Price 2013). Those programmes can be used invest than owners. This effect is found for almost all to encourage the adoption of new technologies such as investment goods studied in this paper, with a sub- energy-efficient appliances and, more generally, stantially larger magnitude for relatively immobile encourage households to engage in energy conserva- investments, such as windows and thermal insulation. tion actions and practices. Labels can be also used to Nevertheless, renters show some propensity to invest provide households with reliable information about in lower-cost technologies that are more mobile, such the performance of energy conservation measures or as energy-efficient appliances and light bulbs. renewable energy, encouraging them to conserve Moreover, investment depends positively on energy and invest. income and this effect is larger for lower income levels. This is indicative of credit constraints. Many energy Acknowledgments efficiency and renewable investments have high initial investment costs representing a relevant obstacle, The research leading to these results has received especially for low-income households, who are more funding from the People Programme (Marie Curie likely to be credit-constrained. Actions) of the European Union’s Seventh Frame- Technology adoption is also influenced by house- work Programme (FP7/2007-2013) under REA holds’ attitudes and beliefs, as households who are in grant agreement PIEF-GA-2012-331154—project an environmental group or who are ready to engage in PACE (Property Assessed Clean Energy). The low-cost energy conservation practices are also more authors would like to thank Walid Oueslati, likely to invest in energy efficiency or renewables. Giuseppe Nicoletti, Ysé Serret, Nick Johnstone, These results suggest that targeted policies are Jérôme Silva, Daniel Kammen and various partici- required to address specific barriers for different pants of OECD seminars for their valuable com- groups of consumers. For instance, credit constraints ments and suggestions. The opinions expressed are more relevant for low-income households and lift- herein are those of the authors and do not ing these constraints would likely promote investment necessarily reflect the opinions of the OECD. for this group. Direct subsidies, tax credits or rebates can also be relevant policy instruments to lower the References upfront cost of energy investments. While internalis- ing external costs of emissions by increasing energy Allcott H 2011 Social norms and energy conservation Journal of prices is thought to be a more efficient instrument in Public Economics 95 1082–95 the absence of credit constraints, subsidies to adopt Attari S, DeKay M, Davidson C and de Bruin W 2010 Public perceptions of energy consumption and savings PNAS Early low-emission technologies may be a more effective Edition 107 16054–9 and less costly than higher energy taxation when credit Bulkeley H and Broto V C 2013 Government by experiment? Global constraints are present. cities and the governing of climate change Trans. Inst. Br. The split incentive problem that arises in the rental Geogr. 38 361–75 Davis L W 2010 Evaluating the slow adoption of energy efficient housing market also requires specific policy actions. If investments: are renters less likely to have energy efficient investments in energy-efficient measures were capita- appliances? National Bureau of Economic Research working lised in the purchase and rental prices of the corre- paper 16114 (www.nber.org/papers/w16114.pdf) sponding property, the owner could recover the Di Maria C, Ferreira S and Lazarova E 2010 Shedding light on the light bulb puzzle: the role of attitudes and perceptions in the investment cost. However, this is often not the adoption of energy efficient light bulbs Scottish J. Political case, not least because the effectiveness of energy effi- Econ. 57 48–67 ciency measures is difficult to observe. Energy effi- Donaldson S and Grant-Vallone E 2002 Understanding self-report ciency labels can help in this respect. In addition, bias in organizational behaviour research J. Bus. Psychol. 17 245–60 explicitly allowing owners to increase the rent after Ferraro P J and Price M K 2013 Using nonpecuniary strategies to implementing energy efficiency measures might be influence behavior: evidence from a large-scale field necessary in some countries, where owners are not experiment Rev. Econ. Stat. 95 64–73 allowed to raise the rent as they wish, unless the tenant Gillingham K, Harding M and Rapson D 2012 Split incentives and household energy consumption Energy J. 33 37–62 changes. In Germany, this seems to have helped to 13 Environ. Res. Lett. 10 (2015) 044015 N Ameli and N Brandt Hassett K A and Metcalf G E 1995 Energy tax credits and residential Morey E, Thacher J and Breffle W 2006 Using angler characteristics conservation investment: evidence from panel data J. Public and attitudinal data to identify environmental preference Econ. 57 201–17 classes: a latent-class model Environ. Res. Econ. 34 IEA 2013 World Energy Outlook (Paris: OECD) (www. 539–65 worldenergyoutlook.org/publications/weo-2013) OECD 2013 Greening household behaviour: overview from the Jaffe A B and Stavins R N 1995 Dynamic incentives of 2011 survey OECD Studies on Environmental Policy and environmental regulations: the effects of alternative policy Household Behaviour (Paris: OECD) (doi:10.1787/ instruments on technology diffusion J. Environ. Econ. 9789264181373-en) Manage. 29 S43–63 Olli E, Grendstad G and Wollebaek D 2001 Correlates of KahnME2007Dogreensdrivehummers orhybrids?environmental environmental behaviour: bringing back social context ideologyasadeterminantof consumer choiceJ.Environ.Econ. Environ. Behav. 33 181–208 Manage. 54 129–45 Poortinga W, Steg L, Vlek C and Wiersma G 2003 Household Kahneman D and Tversky A 1979 Prospect theory: an analysis of preferences for energy-saving measures: a conjoint analysis J. decision under risk Econometrica 47 263–92 Econ. Psychol. 24 49–64 King M F and Bruner G C 2000 Social desirability bias: a neglected Raftery A 1995 Bayesian model averaging selection in social research aspect of validity testing Psychology & Marketing 17 79–103 Soc. Methodol. 25 111–63 Kollmuss A and Agyeman M 2002 Mind the gap: why people act Sardianou E 2007 Estimating energy conservation patterns of Greek environmentally, and what are the barriers to pro- households Energy Policy 35 3778–91 environmental behaviours? Environ. Educ. Res. 8 239–60 Sardianou E and Genoudi P 2013 Which factors affect the Long J 1993 An econometric analysis of residential expenditures on willingness of consumers to adopt renewables energies? energy conservation and renewable energy sources Energy Renew. Energy 57 1–4 Economics 15 232–8 Scott S 1997 Household energy efficiency in Ireland: a replication Mahapatra K and Gustavsson L 2008 An adopter-centric approach study of owner of energy savings items Energy Econ. 19 to analyze the diffusion patterns of innovative residential 187–208 heating systems in Sweden Energy Policy 36 577–90 Shove E 2012 Habits and their creatures The Habits Of Consumption McGuirk P, Bulkeley H and Dowling R 2014 Practices, programs ed A Warde and D Southerton, vol 12 (Helsinki: Collegium) and projects of urban carbon governance: perspectives from pp 100–13 the Australian city Geoforum 52 137–47 Shove E 2014 Putting practise into policy: reconfiguring questions Michelsen C C and Madlener R 2012 Homeowner’s preferences for of consumption and climate change Contemporary Social adopting innovative residential heating system: a discrete Science: J. Acad. Soc. Sci. 9 415–29 choice analysis for Germany Energy Economics 34 1271–83 Simon H A 1959 Theories of decision-making in economics and Mills B and Schleich J 2009 Profits or preferences? assessing the behavioral science Am. Econ. Rev. 49 253–83 adoption of residential solar thermal technologies Energy Sjoberg L and Engelberg E 2005 Lifestyles, and risk perception Policy 37 4145–54 consumer behavior Int. Rev. Sociol.: Revue Internationale de Mills B and Schleich J 2010a What’s driving energy efficient Sociologie 15 327–62 appliances label awareness and purchase propensity? Energy Viklund M 2004 Energy policy options—from the perspective of Policy 38 814–25 public attitudes and risk perceptions Energy Policy 32 Mills B and Schleich J 2010b Why don’t households see the light? 1169–71 Explaining the diffusion of compact fluorescent lamps Willis K, Scarpa R, Gilroy R and Hamza N 2011 Renewable energy Resour. Energy Econ. 32 363–78 adoption in an ageing population: heterogeneity in Mills B and Schleich J 2012 Residential energy-efficient technology preferences for micro-generation technology adoption adoption, energy conservation, knowledge, and attitudes: an Energy Policy 39 6021–9 analysis of european countries Energy Policy 49 616–28 Yates S and Aronson E 1983 A social psychological perspective on Mills B and Schleich J 2014 Household transitions to energy efficient energy conservation in residential buildings Am. Psychol. 38 lighting Energy Economics 46 151–60 435–44

Journal

Environmental Research LettersIOP Publishing

Published: Apr 1, 2015

There are no references for this article.