Residential energy-efficient technology adoption, energy conservation, knowledge, and attitudes: An analysis of European countries

Energy Policy 49 (2012) 616–628

Contents lists available at SciVerse ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Residential energy-efficient technology adoption, energy conservation, knowledge, and attitudes: An analysis of European countries Bradford Mills a,n, Joachim Schleich a,b,c a

Virginia Polytechnic Institute and State University, 314 Hutcheson Hall, Blacksburg 24061-0401, VA, USA Fraunhofer Institute for Systems and Innovation Research, Breslauer Straße 48, 76139 Karlsruhe, Germany c Grenoble Ecole de Management, 12, rue Pierre Se´mard-BP 127, 38003 Grenoble Cedex 01, France b

H I G H L I G H T S c c c c

Household energy use behavior is explored with data from 11 European countries. Household age structure and education influence household energy use behavior and attitudes. Significant country differences in household energy use behavior exist. The EU needs to balance a common energy-efficiency policy framework with country specific policies.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 10 January 2012 Accepted 4 July 2012 Available online 28 July 2012

Relationships between measures of household energy use behavior and household characteristics are estimated using a unique dataset of approximately 5000 households in 10 EU countries and Norway. Family age-composition patterns are found to have a distinct impact on household energy use behavior. Households with young children are more likely to adopt energy-efficient technologies and energy conservation practices and place primary importance on energy savings for environmental reasons. By contrast, households with a high share of elderly members place more importance on financial savings, and have lower levels of technology adoption, energy conservation practice use, and knowledge about household energy use. Education levels also matter, with higher levels associated with energy-efficient technology adoption and energy conservation practice use. Similarly, university education increases the stated importance of energy savings for greenhouse gas reductions and decreases the stated importance for financial reasons. Education impacts also vary greatly across survey countries and there is some evidence of an Eastern–Western European divide with respect to attitudes towards energy savings. These cross-country differences highlight the need to balance a common EU energy-efficiency policy framework with flexibility for country specific policies to address unique constraints to energy-efficient technology and conservation practice adoption. & 2012 Elsevier Ltd. All rights reserved.

Keywords: Household energy-efficiency Technology adoption Energy conservation

1. Introduction As part of the climate and energy package that includes binding 2020 EU27 targets for greenhouse gas emissions and renewable energy use the EU has set an indicative target for energy efficiency (European Commission, 2008a; European Council, 2006, 2007). The EU seems on track to achieve required 20% reductions in greenhouse gas emissions in 2020 compared to 1990 levels, along with 20% renewable energy use in final energy consumption. However, the energy efficiency target of 20% primary energy savings in 2020 compared to business-as-usual

n

Corresponding author. Tel.: þ1 540 231 6461; fax: þ 1 540 231 7417. E-mail address: [email protected] (B. Mills).

0301-4215/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2012.07.008

development will likely be missed without further measures (European Commission, 2008b; Ecofys and Fraunhfoer Institute for Systems and Innovation Research (Fraunhofer ISI) (2010); European Commission, 2011a). To fill the gap, the European Commission proposed a draft Energy Efficiency Directive in June 2011 (European Commission, 2011b). According to the measures contained in the proposal Member States must, among others, implement energy efficiency obligation schemes for utilities (‘‘white certificates’’) to reduce energy sales to final customers, meet annual renovation rate targets for public building floor area, ensure that public bodies only purchase products, buildings and services meeting high energy efficiency standards, or promote energy audits to final customers. Efficiency gains in the household sector, which accounts for about 25% of total final energy consumption and 29% of total

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

electricity use in the EU27 (Bertoldi and Atanasiu, 2009), are also expected to be a key factor in determining whether or not the EU meets its efficiency targets. According to the European Council Action Plan for Energy Efficiency (European Council, 2006) residential energy-savings of 27% may be achieved compared to expected baseline growth by 2020 through the adoption of costefficient residential energy efficient technologies and conservation practices. In a more recent study, Fraunhofer (2009) estimate that the residential sector may cost-effectively save about 19% of final energy compared to the baseline in 2020 with additional policies to overcome barriers to adoption of existing technologies. The bulk of these savings will come from improved thermal insulation, but 7% of energy savings are expected to accrue from the adoption of energy efficient household appliances (including lighting). Additional policy measures to enhance adoption can increase the energy efficient household appliance contribution to final energy savings to about 17% compared to business as usual in 2020. In general, residential energy policies can be employed to both enhance the uptake of improved energy conservation practices (e.g. switching off lights when leaving a room, adjust indoor temperature at night, reduce heat in unused rooms, only use dishwasher and washing machines at full load, put lid on pots) and increase adoption of energy efficient technologies (e.g. insulation of outer walls, attic, window glazing; energy-efficient heating system; purchase energy efficient household appliances, office equipment or light bulbs). The formulation of effective and well targeted residential energy policies to increase both conservation and technology adoption must be based on a sound understanding of how technology adoption, conservation practices, energy use knowledge, and attitudes towards energy conservation are associated with household characteristics. In a diverse regional organization like the EU, it is also essential to identify country-specific differences in energysaving technology adoption and energy conservation practices in order to generate an appropriate combination of common and country-specific policies. This paper employs a unique dataset of almost 5000 households from 11 European countries (10 EU countries and Norway) to identify differences in residential energy efficient technology adoption, energy conservation behavior, and attitudes towards energy savings due to household characteristics and country of residence. Particular emphasis is placed on understanding how education levels may have different impacts on household use across countries. The research is, to our knowledge, the first attempt to analyze residential energy conservation technologies, behavior, and attitudes across a broad cross-section of European countries. The remainder of the paper is organized as follow. After a review of the literature in Section 2, Section 3 lays out the empirical specification of the model. Section 4 provides a description of the data. Results are presented in Section 5 and Section 6 concludes and discusses the main findings and identifies implications for energy policy.

2. Literature overview Household level analyses of the adoption of energy efficient technologies and conservation practices are rather scarce and are concentrated on the US, Canada, and several individual EU countries. Dillman et al. (1983) and Black et al. (1985) examine (primarily thermal) energy efficiency investments and adjustments in behavior using surveys of the Western States of the US and Massachusetts, respectively, while Walsh (1989) and Long (1993) focus on the adoption of thermal energy measures for the entire US. Curtis et al. (1984) analyze technology adoption and

617

behavioral practices aimed at reducing household thermal energy and electricity use in Regina (Canadian Province of Saskatchewan) and Ferguson (1993) analyses the adoption of retrofitting measures for all of Canada. Brechling and Smith (1994) and Caird et al. (2008) explore insulation, heat generation and lighting technologies in UK households. Barr et al. (2005) use data on selected technological measures and conservation practices related to household thermal energy and electricity use for the UK county of Devon. Poortinga et al. (2003, 2004) include an extensive list of technological measures and behavioral practices associated with thermal energy and power use in the Netherlands, while Scott (1997) focuses on several technology measures (attic and hot water cylinder insulation and lighting) in a survey of Irish households. For Germany, Mills and Schleich (2010a) and Mills and Schleich (2010b) explore the adoption of energy-efficient household appliances and of compact fluorescent light bulbs (CFLs), respectively. For Sweden Linden et al. (2006) consider a set of behavioral practices. Mahapatra and Gustavsson (2008) analyze the adoption of heating systems, while Nair et al. (2010) consider several thermal energy investments as well as behavioral practices related to electricity and thermal energy use. Comparing findings across studies and countries is difficult, as studies differ with respect to the types of technologies, behavioral practices, explanatory variables, and methods. Also, comparisons across studies carried out at different points in time may not be warranted, since the technological, social, cultural, economic, and policy environments develop over time. The only cross-country analysis of household adoption of energy efficient technologies and behavioral practices we are aware of is OECD (2011). Respondents in the 10 OECD countries1 included in the studies were found to vary considerably with respect to appliance stock, investments in energy-savings equipment, energy savings behavior, government support received for installations of energyefficient technologies, environmental concerns and attitudes, or motivations to reduce energy consumption. For example, Dutch households are most likely to turn off their electronic appliances and devices. In comparison, households in Australia, the Czech Republic and South Korea are the least likely to switch off appliance in stand-by mode, while households in Sweden and Norway are the least likely to turn off lights when leaving a room. Most studies find that adoption of energy efficient measures and behavioral practices are typically associated with costs (for investments and energy use), habits, and routines, which differ across measures, households and regions. Curtis et al. (1984) were among the first to point out that energy-savings measures may be differentiated as low-cost or no-cost measures which do not involve capital investment but rather behavioral change and high-cost measures which require capital investment and involve technical changes in the residence. From a behavioral perspective it is much easier to change a singular investment decision such as purchasing a CFL than to change daily behavior such as switching off lights when leaving a room (e.g. Gardner, 1996). Also, while energy savings resulting from technology adoption tend to have long run effects, behavioral measures may only have transitory effects (e.g. Abrahamse et al., 2005). Barr et al. (2005) and also OECD (2011) distinguish explicitly between habitual behavior and technology adoption and stress that energy savings behavior needs to be considered within the broader context of environmental behavior. Adoption of energy efficient technologies and conservation measures is usually associated with reduced emissions of greenhouse gases and other pollutants that benefit others without compensating the energy

1 Australia, Canada, the Czech Republic, France, Italy, South Korea, Mexico, the Netherlands, Norway and Sweden.

618

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

savers. In this context, motives for energy savers’ provision of a public good include altruism, empathy, the ‘warm glow of giving’ (Andreoni, 1990) and prestige (Harbaugh, 1998). Studies on the adoption of energy efficient measures in households are typically based on different, partially overlapping concepts from economics (including behavioral economics), psychology and sociology. Insights from the psychology and sociology literature are employed to analyze the impact of psychological variables such as values, beliefs, or attitudes towards energy conservation as well as the impact of social norms shared by relevant groups on energy efficiency activities (Gardner, 1996). The thrust of this literature suggests that attitudes towards energy conservation or environmental motivation in general may at best explain a modest share of the variation in household energy consumption or adoption of ¨ energy savings measures (e.g. Viklund, 2004; Sjoberg and Engelberg, 2005; OECD 2008, 2011; Di Maria et al., 2010). The June 2011 EUROBAROMETER survey (European Commission, 2011c) also hints at differences across EU Member States in terms of environmental concerns and actions. For example, in Luxembourg and Denmark 34% and 31% of the population, respectively, considers climate change to be the single most serious problem facing the world. In Portugal and Ireland, this share is only 7% and 13%, respectively. In the 6 months prior to the survey, about three quarters of Swedes, Slovenes and Luxembourgians report to have taken personal actions to fight climate change, but less than a third of Poles, Romanians or Estonians report the same. Environmental behavior is not only driven by motivational factors, but also determined by contextual factors, including opportunities, individual abilities, status, comfort, and effort (Poortinga et al., 2004, Stern, 2000). In particular, attitudes do not directly determine behavior. Instead they affect intentions which in turn form people’s actions. According to Ajzen and Fishbein (1980, p. 239) intentions are not only influenced by attitudes but also by social pressure and perceived behavioral control. In other words, attitudes towards the environment may not necessarily lead to good intentions, and stated good intentions may not necessarily lead to good actions. Social norms, lack of information about the implications of alternative actions on the environment, or institutional and economic factors may act as barriers towards actual implementation (Van Raaij and Verhallen, 1983). Kammerer (2009) emphasizes the importance of additional customer benefits as a key factor in the demand for ‘‘green’’ products. These additional benefits include energy (and other) cost savings, improved product quality (durability and reliability) or improved repair, upgrade, and disposal possibilities. Households’ information on energy consumption, conservation opportunities and the energy performance of technologies is expected to affect the adoption of energy-efficient technologies. Availability and quality of information about the levels and patterns of current energy consumption depends on the level of metering, the information content of utility bills, and households’ willingness and ability to analyze this information. Similarly, households need to be aware of and able to evaluate energy efficiency opportunities (e.g. Schipper and Hawk, 1991). For example Scott (1997) observes that household knowledge about potential energy savings is associated with higher take-up of energy efficient technologies. Typically, labeling schemes such as those implemented in the EU and US for household appliances are cost-effective measures to overcome barriers related to information and search costs, or to bounded rationality on the part of appliance purchasers (Sutherland, 1991; Howarth et al., 2000). Evaluation studies based on aggregate observed data find that the existing energy labeling programs for household appliances in the US, the EU and Australia are effective in terms of energy and carbon reductions (e.g. Sanchez et al., 2008; Lane et al., 2007; Banerjee and Solomon, 2003; Schiellerup, 2002; Bertoldi,

¨ 1999; Waide, 2001; Waide, 1998). Sammer and Wustenhagen (2006) conduct survey-based conjoint analyses to analyze consumers’ stated choices for washing machines in Switzerland and observe that ecolabeling affects consumers’ purchasing decisions. As for the impact of information campaigns, Reiss and White (2008) observe that consumers respond to both energy prices and information campaigns to reduce energy consumption, although – consistent with the weak correlation between attitude and conservation efforts pointed out above – a survey by the OECD (2008) concludes that information campaigns are not as effective as expected. Households often ignore mass information, but are more likely to respond to well-targeted, direct information (Lutzenhiser, 1993). Similarly, based on stated behavior in Swedish households ¨ the findings by Ek and Soderholm (2010) confirm that providing more concrete information on energy savings measures is likely to be more effective than rather general information. In sum, information may improve the level and the quality of knowledge on energy conservation measures, but improved information need not necessarily result in energy conservation. Based on the empirical literature, factors influencing energy saving activities may generally be categorized as characteristics of the household (education, income, number of children, age, renter or owner), characteristics of the residence (multi-family home, size), characteristics of the measure (behavioral or technological, costs, performance, energy use), economic factors (energy prices), availability and quality of information, weather and climate factors, and attitudes towards energy savings or towards the environment. We will briefly summarize the main findings of the literature, focusing more heavily on factors which are relevant for the subsequent empirical part of the paper.2 2.1. Education Most studies suggest a positive correlation between education level and energy-saving activities, including the econometric analyses by Hirst and Goeltz (1982), Brechling and Smith (1994), Scott (1997) and OECD (2011) for energy efficient technology adoption. Exceptions include Ferguson (1993) and Mills and Schleich (2010a). Among the reasons for a positive correlation are that education reduces the costs of information acquisition (Schultz, 1975). Alternatively, education as a long term investment may be correlated with a low household discount rate and, thus, be positively associated with energy-saving measures that require higher upfront investment costs for energy cost savings that materialize over time. Mills and Schleich (2010a) find that socio-economic factors like higher education levels, higher income, larger households, and higher electricity prices are positively correlated with respondents’ knowledge about the energy efficiency label of appliances. Similarly, Murray and Mills (2011) find in the US household characteristics have a greater impact on EnergyStar label awareness than on the uptake of EnergyStar appliances. Attitudes towards the environment as well as social status, lifestyle (Lutzenhiser, 1992, 1993; Weber and Perrels, 2000), and belonging to a particular social milieu group approving of environment-friendly behavior (e.g. Brand, 1997) also tend to be ˜ as positively related with education. Torgler and Garcı´a-Valin (2007) cite several sources suggesting that higher education levels are associated with higher preferences for environmental conservation. Relationships between education and energy behavior may also differ across countries, in part due to differences in 2 Nair et al. (2010), Brohmann et al. (2009) and Sardianou (2007) include recent surveys of the empirical literature on household energy saving behavior and Wilson and Dowlatabadi (2007) provide a conceptual overview from economics, psychology, sociology and innovation studies.

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

country educational structures (Hofman et al., 2004; Eurostat, 2011). 2.2. Age and household composition The majority of empirical studies analyzing the household uptake of energy efficiency measures and practices control for age (of the household head), but only a few studies account for household composition by age groups. Older household heads may be less likely to adopt energy efficient technologies because the expected rate of return is lower than for households with younger heads. This line of reasoning is supported by the findings of Curtis et al. (1984), Walsh (1989), Poortinga et al. (2003) and Mahapatra and Gustavsson (2008). On the other hand, younger households may be more likely to move and hence be less inclined to invest in energy efficiency improvements, in particular if these measures become an integral part of the built environment. Combining these perspectives, middle aged households should be most likely to adopt capital-intensive energy efficiency measures (e.g. Mills and Schleich, 2010a), particularly if the technologies are structurally linked to the building. For measures with low up-front costs (e.g. light bulbs) and for behavioral measures the expected impact of age is less clear. Lutzenhiser (2002) finds that older households are less likely to adapt behavior while in Mills and Schleich (2010b) adoption intensity of energy efficient light bulbs increases at a declining rate with age. On the other hand, as suggested by CarlssonKanyama et al. (2005), younger households tend to prefer up-todate technology, which is usually also more energy efficient. In sum, the relationship between age and the take-up of energy savings measures is likely to be nonlinear and technology specific. Lower adoption of energy efficient technologies by elder households may also interact with the cohort’s fewer years of formal education, and lower levels of information on energy savings measures. For example, survey results by Linden et al. (2006) for Sweden indicate that younger people have better knowledge about energy-efficient measures than older people. Clustering individuals into different types, the findings by Barr et al. (2005) for the UK, and by Painter et al. (1983) and by Ritchie et al. (1981) for the US suggest that ‘‘energy savers’’ are older. Addressing environmental concerns directly, the studies by Whitehead (1991) and by Carlsson and Johansson-Stenman ˜ as (2007) – found that (2000) – cited by Torgler and Garcı´a-Valin willingness to pay for environmental protection decreases with age, arguably because a shorter expected remaining lifetime results in lower expected benefits from environmental preserva˜ as (2007) for Spain and Torgler et al. tion. Torgler and Garcı´a-Valin (2008) for 33 Western European countries also observe a negative correlation between age and environmental attitudes/preferences. Similarly, according to Howell and Laska (1992) younger people in the US are more concerned about the environment than ˜ as (2007) also older people. However, as Torgler and Garcı´a-Valin point out, age effects need to be decomposed into a life cycle effect which stems from being in a particular stage of life, and into a cohort effect which results from belonging to a particular generation with generation-specific experiences, socialization, and economic conditions (e.g. ‘‘flower power generation’’ versus ‘‘baby boomers’’). Thus, depending on the timing and the region of the survey, age may turn out to have quite different effects on households’ adoption of energy-efficient measures. Young children in the household may also impact adoption, as parents may be more concerned about short and long run local and global environmental effects that will influence current and future wellbeing of their children. Dupont (2004) finds that the number of children is positively related to the adoption of energyefficient technologies and conservation behavior, but Torgler et al.

619

(2008) do not find children to generate a positive shift in parental preferences for environmental conservation.

3. Empirical specification This paper focuses on establishing the empirical relationship between household decision variables (adoption of energy efficient technologies, use of energy conservation practices in the home, knowledge of level of energy use and energy saving options, preferences for energy savings for environmental and for financial reasons) and household characteristics, allowing for country specific effects. Specifically, we econometrically estimate reduced form regression models employing these sets of dependent and independent variables. Country specific education effects are also tested for in alternative specifications for each independent variable. 3.1. Dependent variable measures Household adoption of energy efficient technologies is characterized by two alternative measures. The first measure (buyind) is an index of adoption of energy efficient ‘‘white’’ appliances (refrigerators, freezers, dishwashers, washing machines, and dryers), office equipment, and light bulbs generated by factor analysis. White appliances account for about 25% of residential electricity use in the EU27, lighting for 11%, and computers for about 3% (Bertoldi and Atanasiu 2009, p. 13f). In the EU all major white appliances are classified under a common energy labeling framework from most efficient (class Aþþ) to least efficient (class-G). The index includes a measure of the energy class of the above mentioned major white appliances.3 Many households did not report appliance energy classes, either because the appliance was not owned, was purchased before the rating system was implemented, or because the energy class was not known by the respondent.4 In these cases the energy class is recorded as a zero, equivalent to specifying no contribution to household energy-efficient technologies. However, separate indicator variables are also included in the factor analysis to indicate if cold appliances, washing machines, or tumble dryers without a known rating were purchased in the last five years, as recent purchases are far more likely to have a high energy efficiency rating (Bertoldi and Atanasiu, 2009). Adoption of energy efficient office technologies is measured as the purchase of EnergyStar labeled products. Adoption of the third technology type, energy efficient light bulbs, is simply measured as the share of household bulbs that are energy efficient compact fluorescent bulbs (CFLs). The CFL share of all household bulbs (cflshare) is used as an alternative second measure of energy efficient technology adoption. The sole CFL share measure has the advantage of simplicity. But, by the same token, CFL share is a less comprehensive measure of household adoption of energy efficient technologies. A household knowledge index (knowledge) is also generated through factor analysis. The index is based on three indicators of household knowledge of energy use that are available in the country surveys; if the household knows its annual electricity consumption, if the household correctly knows what the EnergyStar label stands for, and if the household knows that computer monitor screensavers do not save electricity. 3 Cold appliances are coded from G ¼ 1 to Aþþ¼ 9, while dishwashers, washing machines, and tumble dryers are coded from G ¼1 to A ¼7. 4 Implementing directives were published by the EU in 1994 for refrigerators, freezers and their combinations, in 1995 for washing machines, and in 1997 for dishwashers. In 2004, the labeling scheme for cold appliances was extended to A þ and Aþþ to account for substantial energy efficiency improvements in the highest energy efficiency category. Table A1 provides information on the dates that implementation directives became law in specific countries.

620

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

Table 1 Descriptive statistics. Variables Dependent buyind cflshare knowledge effindex goalghe goalsav Independent highschool vocation universtiy lt12 to18 to65 gt65 belgium bulgaria czech denmark france germany greece hungary norway portugal romania Number of observations

Description

Mean

St. dev.

Energy-efficient technology adoption index Share of bulbs that are CFLs Knowledge of energy use and conservation measures index Use of energy conserving practices index Energy savings is most important for greenhouse gas reductions¼ 1 Energy savings is most important for financial reasons ¼1

0.000 0.166 0.000 0.000 0.196 0.633

0.9087

Completed high school, yes ¼1 Completed vocational school, yes ¼ 1 Completed university, yes ¼1 Number of household members less than 12 years of age Number of household members 12–18 years of age Number of household members 19–65 years of age Number of household members greater than 65 years of age Resident of country¼ 1

0.223 0.202 0.508 0.399 0.221 1.973 0.223 0.109 0.103 0.101 0.085 0.020 0.111 0.084 0.100 0.052 0.108 0.127 4915

Similarly, a household energy conservation index (effindex) is generated through factor analysis based on six indicators of energy conservation practices in the home. These practices are (1) fully loading the washing machine every time, (2) cooking frequently with a pressure-cooker, (3) turning off the lights every time a room is vacated, (4) turning off the TV when it is not being watched, (5) setting energy saving features on the computer monitor, and (6) setting energy saving features on the computer desktop. Household attitudes toward energy savings are captured through household indicators of the stated importance of energy savings for environmental (greenhouse gas reduction) reasons (goalghe) and financial reasons (goalsav). Specifically, attitudes are measured by households indicating that they felt it was ‘most important’ to save electricity for that reason. By construction the indexes buyind, knowledge and effindex are normally distributed with zero means, while goalghe and goalsav are dichotomous. 3.2. Independent variables The independent variables employed to establish relationships with the above indexes are driven largely by data availability. Education is measured for the most educated member of the household with no-high-school as the base group and dichotomous indicator variables for high-school, trade or vocational school, and university. Household composition is measured by the number of members less than 12 years of age (lt12), the number of members 13–18 years of age (to18), the number of members 19–65 years of age (to65), and the number of members over 65 years of age (gt65). Country specific effects are captured through country indicators for Belgium, Bulgaria, The Czech Republic, Denmark, France, Greece, Hungary, Norway, Portugal, and Romania, with Germany being the base country. Education indicators are also interacted with country indicators. Resulting parameter estimates can be interpreted as country differences in the impact of the specified education level relative to the same education level in the base country, Germany.

0.2952 0.8008

0.7509 0.5316 1.0159 0.5477

Relationships with continuous indexes are estimated via OLS regression models. Given the large number of observations with a response of zero, relationships with the CFL share of household light bulbs regression are estimated with a Tobit model. Similarly, relationships with the dichotomous environmental attitude indicators are estimated with Probit models.

4. Data The study dataset is generated from the Residential Monitoring to Decrease Energy Use and Carbon Emissions in Europe Project (REMODECE) survey conducted in 11 countries in 2007 (de Almeida et al., 2008). All countries used a common survey instrument that was translated into the local language. The goal was to survey at least 500 households in each country. However, there was considerable variation in country data collection strategies. Belgium, The Czech Republic, Denmark, Norway, and Portugal relied primarily on on-line internet based surveys. Bulgaria and Germany relied primarily on mail surveys, while France used telephone interviews and Hungary and Romania used face-to-face interviews. Greece used a mixture of face-to-face, online, email, and mail surveys.5 Data are downloaded from the project website at: http://www.isr.uc.pt/  remodece/. The overall sample contains 4915 households. The distribution of country sample sizes from the website data ranges from Romania with 622 households to France with 100 households. Descriptive statistics for the variables employed in the study are presented in Table 1. By construction, the means of all the dependent variables generated through factor analysis (buyind, knowledge, effindex) are zero. The cflshare variable indicates that the average share of household bulbs that are CFLs is 16.6%, with 43% of households having no CFL bulbs. For attitudes, 19.6% of households indicated that energy savings was most important for 5 Different modes of survey implementation may influence dependent variable responses. However, it is difficult to disentangle country survey implementation effects from other country fixed effects.

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

621

Table 2 Dependent variable means by country. Variables

Belgium

Bulgaria

Czech

Denmark

France

Germany

Greece

Hungary

Norway

Portugal

Romania

buyind cflshare knowledge effindex goalghe goalsav

0.20 0.20  0.29  0.03 0.15 0.36

0.00 0.11 0.04 0.03 0.10 0.72

0.06 0.25 0.04 0.24 0.20 0.73

0.10 0.17 0.04 0.31 0.45 0.69

0.05 0.17 0.00 0.26 0.40 0.56

 0.25 0.14  0.01  0.51 0.14 0.75

0.07 0.19 0.10 0.39 0.45 0.19

0.10 0.22 0.05  0.39 0.06 0.84

 0.12 0.13 0.07 0.11 0.15 0.61

0.04 0.17 0.00 0.16 0.25 0.59

0.03 0.09 0.03  0.14 0.07 0.76

greenhouse gas reductions and 63.3% indicated that energy savings was most important for financial reasons. Dependent variable means are provided by country in Table 2. Variable means vary significantly across countries. Germany shows the lowest average index of energy efficient technology adoption ( 0.25). Belgium, on the other hand, shows the highest average household index of technology adoption. Average household shares of CFL bulbs show a somewhat different distribution across countries. Romania and Bulgaria show the lowest average shares, while The Czech Republic and Hungary show the highest shares of household bulbs that are CFLs. The knowledge index also differs across countries, with Belgium having the lowest average knowledge index by a significant margin and Greece having the highest average index. Greece also has the highest average household energy conservation index, while Germany (as with the technology index) has the lowest energy conservation index. Stated reasons for the importance of energy conservation also differ greatly across countries. In Denmark and Greece 45% of respondents state that it was ‘more’ important to save energy for greenhouse gas reductions, while in Hungary and Romania only 6% and 7%, respectively, state it was more important for this reason. On the other hand, households in Hungary and Romania were most likely to say it was more important to save electricity for financial savings (84% and 76%, respectively). Households in Greece were least likely (19%) to save electricity for financial savings. Thus, there is considerable variation of all measures of household behavior with respect to energy both across measures and across countries. The underlying reasons for these differences are now analyzed in a multivariate regression framework.

5. Results Two regression model specifications are presented for each measure of household energy behavior. The first specification includes each dependent variable as a linear function of the education indicator variables, the household composition variables, and the country indicators. The second specification interacts each education level indicator with each country indicator in order to estimate country specific effects of education. The interaction term estimates are presented in Tables A1 and A2.6 All regression parameter estimates are evaluated using Huber (1967) and White (1980) robust standard errors, as Breusch and Pagan (1979) tests support the presence of heteroskedastic error terms in all models. Results of the OLS and Tobit models for the energy efficient technology index and the CFL share models, respectively, are presented in Table 3. Findings of OLS regressions for the knowledge index and for the energy conservation index appear in Table 4. Marginal effects from the probit models for the 6

Age-composition and country variable interaction terms are not included in order to retain a parsimonious and solvable model structure.

stated importance of energy savings for environmental reasons and for financial reasons are shown in Table 5.

5.1. Energy efficient technology index (buyind) Column 1 in Table 3 contains parameter estimates for the energy-efficient technology index without education–country interaction terms. Household use of energy-efficient technologies increases when the most educated household member has completed high school, vocational school, and university relative to the non-high-school base. Further, university education impact on energy-efficient technology adoption is larger than highschool or vocational impact. As expected, family structure also influences household energy-efficient technology use. The presence of children less than 12 years old and adults between 19 and 65 years of age is associated with greater energy-efficient technology use, while the presence of household members greater than 65 years of age decreases the expected level of the technology index. All country parameter estimates are positive and statistically significant.7 This result implies that, after controlling for household education levels and age composition, households in all countries have, on average, higher energy-efficient technology indexes than those in Germany, with households in Denmark and Hungary being highest. It is also interesting to note that, with the exception of Belgium, the relative ranking of the parameter estimates corresponds closely to the ranking in the descriptive statistics for the technology index in Table 2. Thus, differences in household education levels and household demographic composition do not appear to be driving observed country differences in energy-efficient technology adoption. Parameter estimates for the energy efficient technology index with country-education level interaction terms are presented in the third column of Table 3. Both country indicators and country– education interaction terms are jointly significant. The education indicator terms by themselves are no longer significant. Since Germany is the benchmark country, this suggests that propensities for households to adopt energy-efficient technologies do not increase with education levels in Germany. The education level impacts in other countries can be recovered from the parameter estimates presented in column 1 of Table A2. Three observations are of note. First, there is a great deal of heterogeneity in education level impacts across countries. Second, for a given country, education impacts tend to be largest and positive at the university level. Third, Hungary and Portugal appear to have the largest impacts of education beyond high school on energyefficient technology adoption. The impacts of family composition on energy-efficient technology adoption remains relatively unchanged with the inclusion 7 Country specific effects are jointly significant in all specifications. Wald tests are employed to test joint significance in all models, except the CFL tobit share models which employ log-likelihood ratio tests.

622

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

Table 3 Regression of technology indexes on household characteristics and countries. OLS estimates buyind

buyind

Parameter Est.

Robust St. Err.

Parameter Est.

highschool vocation university lt12 to18 to65 gt65 belgium bulgaria czech denmark france greece hungary norway portugal romania constant

0.113nn 0.112nn 0.175nn 0.023nn 0.003 0.015nn  0.048nn 0.153nn 0.177nn 0.258nn 0.322nn 0.269nn 0.268nn 0.331nn 0.057nn 0.252nn 0.235nn  0.375nn

0.028 0.028 0.026 0.008 0.011 0.007 0.012 0.027 0.026 0.024 0.023 0.029 0.044 0.026 0.030 0.026 0.023 0.032

 0.138  0.117  0.093 0.024nn 0.005 0.014nn  0.038nn  0.121 0.291nn 0.220n 0.263nn 0.097 0.166  0.043  0.147  0.216nn 0.088  0.144n

Test countries¼0

29.66nn

7.51nn

350.07nn

46.590nn

Includes countryneducation Test countryneducation ¼0

No

Yes 4.21nn

No

Yes 66.70nn

Adj. R2 F-test Log-likelihood LR w2 N Uncensored

0.087 27.400nn

0.1066 13.17nn  2502.5 399.6nn 4915 2786

 2469.1 466.33nn 4915 2786

4915

Tobit estimates cflshare

Tobit estimates cflshare

Robust St. Err.

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

0.086 0.088 0.088 0.008 0.011 0.007 0.012 0.168 0.146 0.120 0.094 0.102 0.147 0.101 0.154 0.094 0.104 0.083

0.100nn 0.103nn 0.137nn 0.021nn 0.008 0.004 0.000 0.047nn  0.176nn 0.145nn 0.056nn 0.031 0.028 0.073nn  0.009 0.019  0.213nn  0.078nn

0.026 0.026 0.024 0.007 0.010 0.006 0.011 0.024 0.026 0.023 0.025 0.025 0.041 0.024 0.029 0.024 0.025 0.030

 0.091  0.092  0.092 0.023nn 0.009 0.003 0.007  0.090  0.060 0.094  0.069  0.111  0.260  0.260nn  0.101  0.284nn  0.585nn 0.113n

0.071 0.074 0.073 0.007 0.010 0.006 0.011 0.156 0.128 0.119 0.095 0.083 0.166 0.088 0.157 0.081 0.128 0.067

4915

Note: Buyind tests of joint significance of county and country-education interaction terms are Wald w2 and cflshare tests are log-likelihood. n

Significant at p ¼ 0.10 level. Significant at p ¼ 0.05 level.

nn

Table 4 OLS regressions of knowledge and conservation indexes on household characteristics and countries. knowledge Parameter Est.

knowledge Robust St. Err.

Parameter Est. nn

effindex Robust St. Err.

Parameter Est. nn

effindex Robust St. Err.

Parameter Est. nn

highschool

0.014

0.015

0.083

0.012

0.174

vocation university lt12 to18 to65 gt65 belgium bulgaria czech denmark france greece hungary norway portugal romania constant

 0.042nn 0.071nn 0.008  0.017n 0.009nn  0.015n  0.275nn  0.002 0.017 0.063nn 0.071nn  0.005 0.055nn 0.040nn  0.027nn 0.002  0.032n

0.018 0.014 0.006 0.009 0.004 0.008 0.027 0.013 0.013 0.014 0.014 0.021 0.012 0.016 0.013 0.012 0.017

0.089nn 0.145nn 0.009  0.014n 0.008nn  0.014n  0.546n 0.099n 0.016 0.209nn 0.222nn 0.010 0.170n 0.111 0.029nn 0.107nn  0.113nn

0.016 0.017 0.006 0.009 0.004 0.008 0.296 0.052 0.127 0.040 0.029 0.009 0.029 0.074 0.013 0.033 0.008

0.168nn 0.409nn 0.026n  0.012 0.029nn  0.114nn 0.386nn 0.348nn 0.628nn 0.780nn 0.772nn 0.683nn 0.083n 0.447nn 0.556nn 0.249nn  0.730nn

Test countries¼0

22.48nn

13.23nn

59.97nn

22.100nn

Includes countryneducation Test countryneducation ¼0

No

Yes 5.39nn

No

Yes 3.17nn

Adj. R2 F-test N

0.148 17.620nn 4915

0.189 24.880nn 4915

0.164 74.370nn 4915

0.178 34.43 4915

Note: Country and country–education interaction are jointly tested with Wald w2. n

Significant at p ¼ 0.10 level. Significant at p ¼ 0.05 level.

nn

0.045

0.274

0.046 0.041 0.015 0.021 0.012 0.021 0.050 0.047 0.044 0.046 0.047 0.081 0.044 0.057 0.045 0.046 0.052

0.368nn 0.469nn 0.026n  0.008 0.026nn  0.102nn 0.160 0.667nn 1.251nn 1.001nn 1.219nn 1.120nn 0.087 0.459 0.366nn 0.446nn  0.839nn

Robust St. Err. 0.076 0.087 0.086 0.015 0.021 0.012 0.022 0.282 0.177 0.165 0.148 0.104 0.307 0.085 0.298 0.097 0.158 0.065

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

623

Table 5 Marginal effects of environmental and cost-savings attitudes on household characteristics and countries. goalghe

highschool vocation university lt12 to18 to65 gt65 belgium bulgaria czech denmark france greece hungary norway portugal romania

goalghe

goalsav

goalsav

Parameter Est.

Stand. Err.

Parameter Est.

Stand. Err.

Parameter Est.

Stand. Err.

Parameter Est.

Stand. Err.

0.041n 0.031 0.067nn 0.007  0.024nn  0.005  0.033nn  0.002  0.081nn 0.037 0.226nn 0.214nn 0.190nn  0.133nn  0.014 0.083nn  0.121nn

0.025 0.025 0.023 0.007 0.011 0.006 0.011 0.023 0.026 0.023 0.022 0.022 0.035 0.027 0.029 0.023 0.025

 0.002  0.065  0.078 0.007  0.022nn  0.005  0.031nn 0.013  0.115  0.124 0.233nn 0.111n 0.089  0.275nn  0.050  0.041  0.219n

0.056 0.052 0.060 0.007 0.011 0.006 0.011 0.039 0.135 0.136 0.075 0.066 0.129 0.111 0.150 0.060 0.118

 0.008  0.037  0.088nn 0.001  0.005 0.001 0.029nn  0.316nn 0.016  0.004  0.048  0.479nn  0.151nn 0.110nn  0.095nn  0.123nn 0.034

0.030 0.030 0.027 0.009 0.012 0.007 0.013 0.027 0.029 0.028 0.030 0.029 0.046 0.030 0.033 0.028 0.027

0.055 0.051  0.033 0.003  0.003 0.000 0.026n  0.390nn  0.054 0.265  0.051  0.409nn  0.021 0.264nn  0.007  0.004  0.027

0.071 0.068 0.073 0.009 0.012 0.007 0.014 0.044 0.137 0.180 0.100 0.087 0.177 0.105 0.176 0.075 0.100

481.03nn

38.44nn

611.94nn

129.610nn

Includes country education Test countryneducation ¼0

No

Yes 41.40nn

No

Yes 53.24nn

Wald w2 Log likelihood N

528.470nn  2150.224 4915

534.680nn  2129.384 4915

699.300nn  2847.255 4915

722.03nn  2820.433 4915

Test countries¼0 n

Note: Belgiumnvocation and Portugalnvocation terms are dropped in goalghe and goalsav regressions. Country and country-education interaction are jointly tested with Wald w2. n

Significant at p¼ 0.10 level. Significant at p ¼ 0.05 level.

nn

of the country–education interaction terms. Country-specific parameter estimates are, however, significantly different. Germany no longer has the lowest base (no high school) average level of energy efficient technology adoption once the low impact of higher education levels in Germany is accounted for. In fact, only Bulgaria and The Czech Republic have significantly higher base country parameter estimates. 5.2. CFL share (cflshare) Turning to the household CFL share parameter estimates in column 5 in Table 3, we see that bulb adoption is higher with a high-school degree or a higher level of education. But the differential increase associated with a university degree appears to be minor. The share of CFL bulbs also increases with the presence of children less than 12 years of age in the household. But, unlike for the technology adoption index, the presence of elderly household members does not lower adoption. Parameter estimates for the country-specific effects again preserve the ranking for CFL household share descriptive statistics seen in Table 2, with Romania lowest, Germany (the benchmark) in the middle, and The Czech Republic having the largest positive country effect.8 As with the full energy-efficient technology index, none of the education level indicators are significant when country-education interaction terms are included in the CFL share model (column 7 in Table 3). The presence of young children in the household still increases the household share of CFL bulbs, but only Hungary, Portugal, and Romania show significant base differences—this implies that households in these countries have lower CFL shares 8

Country effects are jointly significant (p ¼0.05) in a log-likelihood test.

with no high-school than similar households in Germany. However, Hungary, Portugal, and Romania also show positive and significant high-school and university interaction terms. This result suggests that there is a particularly strong increase in CFL shares with highschool and university education in these countries. 5.3. Energy-use-knowledge index (knowledge) Column 1 of Table 4 contains the estimates for the country effects only model parameter estimates. A high-school diploma has no significant impact on household knowledge of energy use relative to no high-school degree. Further, knowledge of household energy use appears to be significantly lower among those with a vocational degree. However, the knowledge index does increase when the most educated member of the household has a university degree. The knowledge index also increases with the presence of adult household members age 19–65, but decreases with the presence of household members older than 65 and between 12 and 18 years of age. Taken as a whole the household age composition results suggest that adult families with young children have the highest level of knowledge of household energy use and energy savings possibilities. In terms of country effects, there is little correspondence with the estimates for the technology index. Households in Belgium have on average the lowest knowledge index after controlling for household education and age composition, followed by Portugal. A number of countries (Bulgaria, The Czech Republic, Greece, and Romania) have country knowledge index effects that are not significantly different than the base country Germany, while France, Denmark, Hungary, and Norway have significantly higher country knowledge indexes. Column 3 of Table 4 presents parameter estimates for the knowledge index model with country–education interaction

624

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

terms. The base high-school, vocational school, and university educational degree indicators are now positive and significant. Again, this indicates that households in Germany have higher knowledge indexes at higher education levels, particularly at the university degree level. The relationship between the knowledge index and education appears to be particularly strong in Germany, as the significant interaction terms for other countries are all negative. Again, the general result gathered from the country-education interaction terms is that there is a great deal of heterogeneity in educational impacts by country. 5.4. Energy conservation index (effindex) Parameter estimates for the household conservation index without country-education interaction terms are presented in column 5 of Table 4. The parameter estimates suggest education has a strong impact on household energy conservation. Highschool and vocational degrees for the most educated household member have approximately equal impacts on the energy conservation index, while a university degree has more than double the impact of high-school and vocational degrees in terms of increasing the energy conservation index. Household age group composition impacts on the energy conservation index are similar to those estimated for the technology and knowledge indexes; the presence of young children under 12 years of age increases energy conservation behavior in the household, as does the presence of adults 19–65 years of age. But the energy conservation index decreases with the presence of elderly household members over 65 years of age. Country effects are notable in that, as in the descriptive statistics, Germany shows a significantly lower index than all other countries. The relative ranking of country effects in the descriptive statistics in Table 2 is also preserved, with The Czech Republic, Denmark, France, and Greece showing the highest country effects relative to the base, Germany. When country–education interaction terms are included in the conservation model a particularly strong education effect is found for the base country, Germany.9 Thus German households where no adult member has a high school degree show very low propensities to employ energy conservation measures, but propensities increase rapidly with household education levels. Further, most significant interaction terms for other countries are negative, suggesting the impacts of education on household energy conservation are not as strong in these countries as in Germany. However, Portugal is a notable exception, where the interaction terms for both high-school and university are positive and significant. 5.5. Importance of energy savings for environmental reasons (goalghe) Estimates for the stated importance of saving electricity in order to reduce greenhouse gases are presented in column 1 of Table 5. Households where the most educated adult member has a high-school degree and a university degree are 4% and 7% more likely, respectively, to attach primary importance to saving electricity for greenhouse gas reductions than households without an adult with a high-school degree. In terms of household age composition, the presence of children 12–18 years of age and adults greater than 65 years of age both decrease the likelihood of the household stating saving electricity of greenhouse gas reductions was of primary importance. Taken together the results 9 Belgium-vocation and Portugal-vocation terms were dropped in the environmental and cost-savings attitudes regression equations due to convergence problems stemming from the small number of households in this education grouping.

suggest families with young children are most likely to be motivated to saving energy for greenhouse gas reductions. Significant differences in country propensities are also found that closely correspond to the country differences in the descriptive statistics in Table 2. Hungarian, Romanian, and Bulgarian households are less likely than German households to state that it is ‘more’ important to save electricity for greenhouse gas reductions. On the other hand, Denmark, France, Greece, and Portugal are more likely to state it is ‘more’ important to save electricity for greenhouse gas reductions. Marginal effects for the greenhouse gas reductions model with education–country interaction terms is presented in column 3 of Table 5. The base educational affect-now referring to German households with no adult member with a high-school education disappear. Base country affects are also slightly muted, with only Denmark and France showing positive marginal effects and Hungary and Romania showing negative effects relative to Germany. Again, these marginal effects are indicative of country differences among households with no adult member with a high-school education. However, education does appear to have a significant positive impact in several countries. Notably, Greece, Hungary, and Portugal all have significant positive interaction terms for university education. Family age-composition impacts remain essentially unchanged from those in the specific without interaction terms. 5.6. Importance of energy savings for financial reasons (goalsav) Marginal effects in the model of importance of electricity savings for financial reasons appear in column 5 of Table 5 and are very different from those estimated in the greenhouse gas model. University education significantly decreases the probability that the household indicates it is ‘more’ important to save electricity for financial reasons, while adult household members 65 years of age or older increase the probability of this response. In terms of country effects, Germany, Bulgaria, The Czech Republic, and Denmark appear to be the most motivated to save electricity for financial reasons, as the other countries have significantly negative parameter estimates. French households show the largest negative effect, and are 48 percentage points less likely to state that it is ‘more’ important to save electricity for financial reasons. When country–education interaction terms are added to the financial savings model, the negative university parameter estimate is no longer significant. However, the positive parameter estimate for adult household members 65 years of age or older remains with the inclusion of country–education interaction terms. Country effects are, however, more muted. Marginal effects for Belgium and France remain negative, indicating lower stated preferences to save for financial reasons in countries among the least educated households who do not have an adult member with at least a high-school degree. On the other hand, the marginal effect for Hungary is positive. Country–education interaction terms are, as in all other models, jointly significant. But few country– education interaction term marginal effects are significant. Specifically, Czech-Vocational and Hungary-University effects are negative, while the Belgium-University effect is positive.

6. Discussion and conclusions The regression models employed in this analysis contain only a limited sub-set of the variables that are likely to influence household behavior with respect to energy. A crucial missing variable is actual household energy expenditures. Further research is needed to identify how attitudes and behaviors with respect to energy use result in actual changes in energy expenditures, as well as how high and low

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

energy households differ with respect to attitudes and use of energyefficient technologies and conservation practices. However, even the limited set of variables in the REMODECE survey generates important energy policy implications for Europe. The first, and most consistent, result is that there are distinct family age cohort patterns with respect to household energy behavior. Families with young children are more likely to adopt energy-efficient technologies and use energy conservation practices in the home. Conversely, elderly household members make households less likely to adopt technologies and conservation practices. Elderly members also appear to lower household levels of knowledge with respect to energy use. Age cohort differences also generate differences in household attitudes with respect to electricity savings. Families with a high share of elderly members are more likely to attach primary importance to saving electricity for financial reasons and less importance to saving electricity for greenhouse gas reductions. The family-age-cohort results have two implications for household energy efficiency policies. First, it may be difficult to increase energy efficiency technology adoption and energy conservation practice use among the elderly households, as constraints occur in terms of behavior, knowledge, and attitudes. Related, the elderly often have a more limited time horizon over which they make decisions and are particularly unlikely to make long-term investments in energy-saving technologies. Aggregate energy efficiency will occur as the current generation of elderly household heads cease to be primary decision makers with respect to household energy behavior, albeit slowly. Second, some openings do exist for increased energy efficiency among the current elder segment of the population. In particular, households with elderly members do appear to be motivated to save energy for financial reasons. So technologies and conservation practices that clearly demonstrate cost-effective energy savings in the short-run are likely to gain purchase among this segment of the population. Information dissemination effects that stress financial savings associated with simple conservation efforts and technology investments with relatively short breakeven points may be particularly effective among the elderly. Appliance energy-efficiency labels can also be modified to include information on electricity operating costs as well as electricity use and on expected break-even points compared to lower initial capital cost, but less energy efficient, models. Elderly households may also be responsive to combinations of feedback on energy use and dynamic electricity pricing policy options being explored by the EU (e.g. regulation on smart metering) that make the financial implications of household energy use decisions clear. However, this interest in cost savings is offset by the reluctance among elderly to adopt technologies needed to provide residential feedback on energy use. Education impacts show greater variation across household energy behavior, knowledge, and attitudes. Education levels show strong positive impacts on household energy-efficient technology adoption and on household use of energy conservation practices. The relationship between education levels and household knowledge of energy use and energy savings options is less consistently strong. This suggests education may be more effective in fostering actual behavioral changes than in increasing knowledge of energy saving opportunities. Attitudes also change with education, with stated motivations shifting toward energy savings for environmental reasons at higher education levels. As with the elderly households, low education households are mainly motivated by financial savings to save energy. Yet, low education households are also less likely to adopt energy savings technologies. Combined, these results suggest that low education households either find available energy-efficient technologies to not be profitable, lack information on profitability, face

625

credit-constraints to adopting capital intensive technologies, or are more likely to rent rather than own large appliances. Further empirical research is needed to better identify the nature of constraints to energy-efficient technology adoption and conservation and generate policies to address specific constraints. Lack of adoption of highly profitable conservation measures in low education households may, as with the elderly, stem from information constraints, which can be addressed through targeted advertising campaigns, perhaps at application points for social assistance programs. Research results also highlight a great deal of country heterogeneity in household energy-efficient technology adoption, household use of energy conservation practices, and household attitudes towards energy savings. Some of this heterogeneity can be partitioned into an East–West divide with respect to attitudes towards energy savings. Specifically, propensities to attach primary importance to electricity savings for greenhouse gas reductions are lower in Eastern Europe (Bulgaria, Hungary, and Romania), while propensities to attach primary importance to savings for financial reasons are high in these same countries. Household income, which is not available in the dataset, may also play a role in observed crosscounty differences. For instance, the household energy use knowledge index is not only lower in Eastern European countries, but also in some of the lower income countries in Western Europe (Greece and Portugal). The significance of country–education interaction terms also make clear that educational impacts on household energy use behavior vary across countries even after controlling for specific attributes. The results have important implications for energy-efficiency policy both for the EU as a whole and for individual Member States. The documented cross-country heterogeneity highlights the need for a framework that provides common country goals and targets, but remains flexible enough to address country specific constraints to the household adoption of energy conservation practices and energy efficient technologies. Examples of country-specific measures within a broad EU energy policy framework include financial instruments, information campaigns, and national targets. In addition to complying with the EU appliance labeling directive, Members States may generate financial instruments to subsidize (via discount vouchers, rebates, and tax reductions) replacement of old appliance with new energy-efficient units (as in Austria, Portugal, Denmark, Brussels region in Belgium, Hungary and the Netherlands).10 Other possible instruments include subsidies and/ or promotion of energy audit programs in general (e.g. Portugal) or specifically for low-income (low education) households (e.g. pilot projects in Germany). Other potential financial instruments include reduced value added taxes for energy-efficient technologies (which have been discussed in France and UK, but was not implemented), and credit provision programs for the purchase of energy efficient appliances (e.g. Bulgaria for cooling systems and heat pumps). A key to effective financial instruments is designing them to induce marginal adopters (rather than subsidizing those who would have adopted anyway). Thus, to the extent possible, financial instruments should be targeted to low education and low income households, and to elderly households. However, such conditionalities for financial incentive programs based on household characteristics are often difficult to implement. Interventions might also be targeted to specific energy consumption levels, particularly if levels are strongly correlated with household characteristics. 10 Examples of energy efficiency policies in particular Member States were primarily taken from Deliverable 11 (Policy Recommendations) within the REMODECE project (de Almeida 2008) and downloaded from www.isr.uc.pt/  REMODECE).

626

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

Country efforts to improve information and raise awareness about energy efficient technologies and conservation practices include mobile ‘‘Energy-Buses’’ to disseminate energy efficiency appliances and technologies among the population in general or within low-income/low education districts (e.g. in Germany Energy-Buses systematically visit smaller and medium-sized communities). Retailers can also be trained to better address the information constraints and specific needs of elder and lower education customers, particularly with respect to label implications for overall energy costs. Romania had a program to train retailers, but the program did not specifically target old/loweducation customers.

Given country differences, it might be most effective if the Energy Efficiency Directive (European Commission, 2011b) sets national targets rather than prescribing particular measures for EU Member States to implement. States would then have the freedom to design their own energy-efficiency policies to meet targets within their unique conditions and constraints. Arguably, a more flexible approach will assist countries to meet 2020 energy-efficiency targets in the most cost effective manner.

Acknowledgement Table A1 Year of country implementation of EU energy consumption labeling directives. Source: MURE2 database.

Belgium Bulgaria Czech Republic Denmark France Germany Greece Hungry Norway Portugal Romania

Refrigerators and freezers

Washing machines

Dishwashers

1999 2006 2004 1995 1995 1998 1996 2002 1996 1995 2001

1999 2006 2004 1996 1996 1998 1997 2002 1996 1996 2001

1999 2006 2004 1999 1998 1998 1997 2002 1996 2000 2001

We are thankful for the helpful comments provided by two anonymous reviewers. We also gratefully acknowledge funding within the project ‘‘Social, ecologic and economic dimensions of ¨ sustainable energy consumption’’ (Soziale, okologische und ¨ okonomische Dimensionen eines nachhaltigen Energiekonsums ¨ in Wohngebauden) under the research program ‘‘From words to deeds—new ways of sustainable consumption’’ (Vom Wissen zum Handeln–neue Wege zum nachhaltigen Konsum) of the German Ministry of Education and Research.

Appendix For the interaction term estimates, see Tables A2.

Table A2 Education and countries interaction parameters. OLS estimates buyind

cflshare

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

Parameter Est.

Robust St. Err.

0.185 0.162 0.126 0.107 0.130 0.183 0.111 0.163 0.105 0.111 0.173 0.161 0.132 0.101 0.123 0.170 0.111 0.225 0.137 0.123 0.174 0.150 0.126 0.102 0.110 0.159 0.112 0.160 0.101 0.110

0.171  0.106 0.036 0.096 0.086 0.498nn 0.343nn 0.174 0.297nn 0.330nn 0.163  0.127 0.060 0.123 0.123 0.289 0.341nn 0.173 0.153 0.281n 0.139  0.085 0.094 0.171n 0.183nn 0.266 0.398nn 0.093 0.373nn 0.448nn

0.167 0.143 0.125 0.113 0.109 0.189 0.097 0.165 0.092 0.135 0.161 0.156 0.133 0.103 0.103 0.186 0.099 0.198 0.147 0.153 0.161 0.133 0.125 0.105nn 0.091 0.177 0.098 0.162 0.089 0.133

0.206  0.028 0.010  0.179nn  0.161nn 0.030  0.109nn  0.072 0.019  0.085nn 0.102  0.141nn 0.029  0.164nn  0.064  0.014  0.140nn  0.015  0.002  0.127nn 0.469  0.10n2  0.002  0.208  0.183nn  0.033  0.132nn  0.065  0.066nn  0.107nn

0.312 0.061 0.127 0.052 0.048 0.051 0.036 0.079 0.026nn 0.037 0.299 0.061 0.130 0.045 0.044 0.043 0.037 0.104 0.058 0.046 0.297 0.055 0.128 0.048 0.036 0.033 0.038 0.078 0.024 0.038

0.339  0.114  0.682nn  0.186  0.517nn  0.371 0.016 0.117 0.278  0.276 0.126  0.220  0.577  0.321n  0.622nn  0.305  0.185  0.613  0.229  0.231 0.282  0.307  0.573nn  0.182  0.436nn  0.544n 0.151 0.029 0.305nn  0.115

0.313n 0.203 0.178 0.189 0.167 0.366 0.115 0.315 0.129 0.175 0.293 0.245 0.197 0.166 0.162 0.342 0.121 0.394 0.247 0.217 0.294 0.189 0.179 0.168 0.126 0.328 0.122 0.309 0.120 0.175

 0.159  0.002 0.113  0.104 0.127n 0.042 0.058 0.037 0.071 0.070

0.083 0.151 0.140 0.093 0.087 0.152 0.122 0.158 0.072 0.125

0.055  0.022  0.297 0.087 0.053 0.008  0.156  0.094  0.091 0.042

0.082 0.153 0.185 0.128 0.117 0.210 0.117 0.186 0.090 0.110

0.056 0.269n 0.006 0.136n 0.126 0.033 0.038

0.161 0.143 0.079 0.079 0.145 0.130 0.204

 0.073  0.344n  0.018  0.023  0.096  0.085  0.185

0.161 0.190 0.106 0.108 0.199 0.118 0.223

0.195 0.068 0.099 0.216 0.056 0.159nn 0.170 0.287nn 0.088 0.204nn 0.153

0.136 0.056 0.140 0.141 0.087 0.076 0.139 0.119 0.156 0.070 0.125

0.113 0.160nn 0.099  0.241  0.010  0.098  0.193  0.224nn  0.077  0.135 0.080

0.130 0.064 0.143 0.185 0.113 0.099 0.189 0.116 0.182 0.086 0.109

belgiumhs 0.315n bulgariahs  0.074 czechhs 0.066 denmarkhs 0.096 greecehs 0.137 francehs 0.307n hungaryhs 0.419nn norwayhs 0.290n portugalhs 0.480nn romaniahs 0.103 belgiumvoc 0.263 bulgariavoc  0.203 czechvoc 0.030 denmarkvoc 0.056 greecevoc 0.068 francevoc 0.190 hungaryvoc 0.411n norwayvoc 0.086 portugalvoc 0.380nn romaniavoc 0.181 belgiumunv 0.312n bulgariaunv  0.086 czechunv 0.058 denmarkunv 0.036 greeceunv 0.222nn franceunv 0.019 hungaryunv 0.395nn norwayunv 0.220 portugalunv 0.568nn romaniaunv 0.198n n

Significant at p¼ 0.10 level. Significant at p ¼ 0.05 level.

nn

knowledge

effindex

goalghe

goalsav

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

References Abrahamse, W., Steg, L., Vlek, C., Rothengatter, T., 2005. A review of intervention studies aimed at household energy conservation. Journal of Environmental Psychology 25, 273–291. Ajzen, I., Fishbein, M. (Eds.), 1980. Understanding Attitudes and Predicting Social Behaviour. Prentice-Hall, New Jersey. Andreoni, J., 1990. Impure altruism and donations to public goods: a theory of warm-glow giving. Economic Journal 100, 464–477. Banerjee, A., Solomon, B.D., 2003. Eco-labeling for energy efficiency and sustainability: a meta-evaluation of the US programs. Energy Policy 31, 109–123. Barr, S., Gilg, A.W., Ford, N., 2005. The household energy gap: examining the divide between habitual- and purchase-related conservation behaviours. Energy Policy 33, 1425–1444. Bertoldi, P., 1999. Energy efficient equipment within SAVE: activities, strategies, success and barriers. In: Proceedings of the SAVE Conference for an Energy Efficient Millennium. Graz, Austria, 8–10 November. /www.eva.wsr.ac.atS. Bertoldi, P., Atanasiu, B., 2009. Electricity Consumption and Efficiency Trends in the Enlarged European Union. Status Report 2009. Black, J.S., Stern, P.C., Elworth, J.T., 1985. Personal and contextual influences on household energy adaptations. Journal of Applied Psychology 70, 3–21. Brand, K.-W., 1997. Environmental consciousness and behaviour: the greening of lifestyles. In: Redclift, M., Woodgate, G. (Eds.), The International Handbook of Environmental Sociology. Edward Elgar, Cheltenham, pp. 204–217. Brechling, V., Smith, S., 1994. Household energy efficiency in the UK. Fiscal Studies 15, 44–56. Breusch, T.S., Pagan, A.R., 1979. A simple test for heteroscedasticity and random coefficient variation. Econometrica 47, 1287–1294. ¨ Brohmann, B., Heinzle, S., Rennings, K., Schleich, J., Wustenhagen, R., 2009. What’s Driving Sustainable Energy Consumption? A Survey of the Empirical Literature, ZEW Discussion Paper No. 09-013, Mannheim et al. Caird, S., Roy, R., Herring, 2008. Improving the energy performance of UK households: results from surveys of consumer adoption and use of low- and zerocarbon technologies. Energy Efficiency 1, 149–166. Carlsson, F., Johansson-Stenman, O., 2000. Willingness to pay for improved air quality in Sweden. Applied Economics 32, 661–669. Carlsson-Kanyama, A., Linden, A.L., Eriksson, B., 2005. Residential energy behavior: does generation matter? International Journal of Consumer Studies 29, 239–252. Curtis, F., Simpson-Housley, P., Drever, S., 1984. Household energy conservation. Energy Policy 12, 452–456. de Almeida, A., et al., 2008. Residential Monitoring to Decrease Energy Use and Carbon Emissions in Europe (REMODECE), Publishable Report. Supported by Intelligent Energy Europe. ISR-University of Coimbra. /www.isr.uc.pt/  remodeceS. Dillman, D., Rosa, E., Dillman, J., 1983. Lifestyle and home energy conservation in the United States: the poor accept lifestyle cutbacks while the wealthy invest in conservation. Journal of Economic Psychology 3, 299–315. Di Maria, C., Ferreira, S., Lazarova, E., 2010. Shedding light on the light bulb puzzle: the role of attitudes and perceptions in the adoption of energy efficient light bulbs. Scottish Journal of Political Economy 57, 48–67. Dupont, D.P., 2004. Do children matter? An examination of gender differences in environmental valuation. Ecological Economics 49, 273–286. Ecofys and Fraunhfoer Institute for Systems and Innovation Research (Fraunhofer ISI), 2010. Energy Savings 2020. How to Triple the Impact of Energy Saving Policies in Europe. Report Commissioned by the European Climate Foundation (ECF) and the Regulatory Assistance Project (RAP). Uetrecht and Karlsruhe. ¨ Ek, K., Soderholm, P., 2010. The devil is in the details: household electricity saving behavior and the role of information. Energy Policy 38, 1578–1587. European Commission, 2008a. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions—20-20-20 by 2020—Europe’s climate change opportunity, COM(2008) 13 final. European Commission, 2008b. Communication from the Commission—energy efficiency: delivering the 20% target. COM(2008) 772 final. European Commission, 2011a. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions—Energy Efficiency Plan 2011. European Commission, 2011b. Proposal for a Directive of the European Parliament and of the Council on Energy Efficiency and Repealing Directives 2004/8/EC and 2006/32/EC. COM/2011/0370 final—COD 2011/0172. European Commission, 2011c. Special Eurobarometer 372. Climate Change. European Council, 2006. Presidency Conclusions of 23/24 March 2006. 7775/1/06 REV1 of 18.05.2006. European Council, 2007. Presidency Conclusions of the European Council of 8/9 March 2007. 7224/1/07 REV 1. Eurostat, 2011. Labour Market Statistics. Eurostat Pocketbooks. Luxembourg: Publications Office of the European Union, 2011. Ferguson, M.R., 1993. Energy-saving housing improvements in Canada (1979–82): a nested logit analysis. Environment and Planning A 25, 609–625. Fraunhofer-Institute for Systems and Innovation Research (Fraunhofer ISI), ENERDATA, Institute of Studies for the Integration of Systems ISIS. Technical University Vienna, and Wuppertal Institute for Climate, Environment and Energy WI (2009). Study on the Energy Savings Potentials in EU Member States, Candidate Countries and EEA Countries. Final Report for the European

627

Commission Directorate-General Energy and Transport. EC Service Contract Number TREN/D1/239-2006/S07.66640. Karlsruhe/Grenoble/Rome/Vienna/ Wuppertal, 15 March 2009, revised. /http://ec.europa.eu/energy/efficiency/ studies/efficiency_en.htmS. Gardner, Stern, 1996. Gardner, Environmental Problems and Human Behaviour. Allyn and Bacon, Boston. Harbaugh, W.T., 1998. The prestige motive for making charitable transfers. American Economic Review 88, 277–282. Hirst, E., Goeltz, R., 1982. Residential energy conservation actions: analysis of disaggregated data. Energy Systems and Policy 6, 135–150. Hofman, R., Hofman, A., Gray, J., Daly, P., 2004. Institutional Contexts of Education Systems in Europe, A Cross-country Comparison of Quality and Equity. Kluwer Academic Press. Howarth, R.B., Haddad, B.M., Paton, B., 2000. The economics of energy efficiency: insights from voluntary participation programmes. Energy Policy 28, 477–486. Howell, S.E., Laska, L.B., 1992. The changing face of the environmental coalition: a research note. Environment and Behavior 24, 134–144. Huber, P., 1967. The behavior of the maximum likelihood estimates under nonstandard conditions. In: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, vol. 1, University of California Press, Berkeley, pp. 221–233. Kammerer, D., 2009. The effects of customer benefit and regulation on environmental product innovation. Empirical evidence from appliance manufacturers in Germany. Ecological Economics, 2285–2295. Lane, K., Harrington, L., Ryan, P., 2007. Evaluating the impact of energy labelling and MEPS—a retrospective look at the case of refrigerators in the UK and Australia. In: European Council for Energy–Efficient Economy (Paris): Proceedings of the 2007 eceee Summer Study. Saving energy—just do it! La Colle sur Loup, Cˆote d’Azur, France, 4–9 June 2007, pp. 743–751. Linden, A-L., Carlsson-Kanyama, A., Eriksson, B., 2006. Efficient and inefficient aspects of residential energy behaviour: what are the policy instruments for change? Energy Policy 34, 1918–1927. Long, J., 1993. An econometric analysis of residential expenditures on energy conservation and renewable energy sources. Energy Economics 15, 232–238. Lutzenhiser, L., 1992. A cultural model of the household energy consumption. Energy 17, 47–60. Lutzenhiser, L., 1993. Social and behavioral aspects of energy use. Annual Review of Energy and Environment 18, 247–289. Lutzenhiser, L., 2002. An Exploratory Analysis of Residential Electricity Conservation Survey and Billing Data, Southern California Edison, Consultant Report, California Energy Commission, May. Mahapatra, K., Gustavsson, L., 2008. An adopter-centric approach to analyze the diffusion patterns of innovative residential heating systems in Sweden. Energy Policy 36, 577–590. Mills, B., Schleich, J., 2010a. What’s driving energy efficient appliance label awareness and purchase propensity? Energy Policy 38, 814–825. Mills, B., Schleich, J., 2010b. Why don’t households see the light? Explaining the diffusion of compact fluorescent lamps. Resource and Energy Economics 32, 363–378. Murray, A., Mills, B., 2011. Read the label! Energystar appliance label awareness and uptake among U.S. consumers. Energy Economics 33, 1103–1110. Nair, G., Gustavsson, L., Mahapatra, K., 2010. Factors influencing energy efficiency investments in existing Swedish residential buildings. Energy Policy 38, 2956–2963. OECD, 2008. Household Behaviour and the Environment. Reviewing the Evidence. OECD. Paris. OECD, 2011. Greening Household Behaviour. The Role of Public Policy. OECD Paris. Painter, J., Semenik, R., Belk, R., 1983. Is there a generalized conservation ethic? A comparison of the determinants of gasoline and home heating energy conservation. Journal of Economic Psychology 3, 317–331. Poortinga, W., Steg, L., Vlek, C., Wiersma, G., 2003. Household preferences for energy-saving measures: a conjoint analysis. Journal of Economic Psychology 24, 49–64. Poortinga, W., Steg, L., Vlek, C., 2004. Values, environmental concern, and environmental behavior. A study into household energy use. Environment and Behavior 36, 70–93. Reiss, P.C., White, M.W., 2008. What changes energy consumption? Process and public pressures. RAND Journal of Economics 39, 636–663. Ritchie, J.R.B., McGougall, G.H.G., Claxton, J.D., 1981. Complexities of household energy consumption and conservation. Journal of Consumer Research 8, 233–242. ¨ Sammer, K., Wustenhagen, R., 2006. The influence of eco-labelling on consumer behaviour—results of a discrete choice analysis for washing machines. Business Strategy and the Environment 15, 185–199. Sanchez, M.C., Brown, R.E., Webber, C., Homan, G.K., 2008. Savings estimates for the United States Environmental Protection Agency’s ENERGY STAR voluntary product labeling program. Energy Policy 36, 2098–2108. Sardianou, E., 2007. Estimating energy conservation patterns of Greek households. Energy Policy 35, 3778–3791. Schiellerup, P., 2002. An examination of the effectiveness of the EU minimum standard on cold appliances: the British case. Energy Policy 30, 327–332. Schipper, L., Hawk, D., 1991. More efficient household electricity use. Energy Policy 19, 244–265. Schultz, T.W., 1975. The value of the ability to deal with disequilibrium. Journal of Economic Literature 13, 827–846.

628

B. Mills, J. Schleich / Energy Policy 49 (2012) 616–628

Scott, S., 1997. Household energy efficiency in Ireland: a replication study of owner of energy saving items. Energy Economics 19, 187–208. ¨ Sjoberg, L., Engelberg, E., 2005. Lifestyles, and risk perception consumer behavior. International Review of Sociology/Revue Internationale de Sociologie 15, 327–362. Stern, P.C., 2000. Toward a coherent theory of environmentally significant behavior. Journal of Social 56, 407–424. Sutherland, R.J., 1991. Market barriers to energy efficiency investments. The Energy Journal 12, 15–34. ˜ as, M.A., Macintyre, A., 2008. Differences in Preferences Torgler, B., Garcı´a-Valin Towards the Environment: The Impact of a Gender, Age and Parental Effect. Fondazione Eni Enrico Mattei, Nota di lavorno: 18.2008. ˜ as, M.A., 2007. The determinants of individuals’ attitudes Torgler, B., Garcı´a-Valin towards preventing environmental damage. Ecological Economics 63, 536–552. Van Raaij, F., Verhallen, T., 1983. A behavioral model of residential energy use. Journal of Economic Psychology 1, 39–63. Viklund, M., 2004. Energy policy options—from the perspective of public attitudes and risk perceptions. Energy Policy 32, 1169–1171.

Waide, P., 1998. Monitoring of Energy Efficiency Trends of European Domestic Refrigeration Appliances, Final Report. PW Consulting for ADEME on behalf of the European Commission (SAVE). PW Consulting: Manchester; 1998. Waide, P., 2001. Monitoring of Energy Efficiency Trends of Refrigerators, Freezers, Washing Machines and Washer-driers Sold in the EU, Final Report. PW Consulting for ADEME on Behalf of the European Commission (SAVE). PW Consulting, Manchester, 2001. Walsh, M., 1989. Energy tax credits and housing improvement. Energy Economics 11, 275–284. Weber, C., Perrels, A., 2000. Modelling lifestyle effects on energy demand and related emissions. Energy Policy 28, 549–566. White, H., 1980. A heteroskedasticity-consistent covariance-matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817–838. Whitehead, J.C., 1991. Environmental interest group behaviour and self-selection bias in contingent valuation mail surveys. Growth and Change 22, 10–21. Wilson, C., Dowlatabadi, H., 2007. Models of decision making and residential energy use. Annual review of environment and resources 32, 169–203.