May 2017 – Briskee, Deliverables & reports
[D 2.3 Scientific working paper on energy efficiency technology adoption (Final 19 April 2018)]
Determinants of households adoption of energy efficient technologies in Europe: focussing on preferences for risk, time and losses
Authors: Joachim Schleich, Xavier Gassmann, Corinne Faure, Thomas Meissner
This paper empirically studies the relation between household adoption of EETs (LEDs, energy efficient appliances, retrofit measures) and risk aversion, standard time preferences, present bias, and loss aversion. The analysis relies on a large representative sample drawn from eight EU countries. Time, risk, and loss aversion preferences were elicited and jointly estimated from participant choices in context-free MPLs.
Improving energy efficiency is commonly considered to be the cheapest short- to medium-term option for meeting energy and climate targets (e.g. IEA 2016), and it is a prime policy goal in many countries. For example, the European Union aims to reduce energy use by at least 30 percent compared to the projected use of energy in 2020. To address environmental externalities such as global warming or resource use, governments employ policies such as technology standards, information measures (e.g. labeling), rebates, tax credits, or subsidized loans. In addition, well designed policies may also help overcome the so-called “energy efficiency paradox”, according to which decision-makers may fail to invest in energy-efficient technologies even though these appear to pay off under prevailing market conditions (e.g. Jaffe and Stavins 1994, Sorrell et al. 2004, Allcott and Greenstone 2012, Gillingham and Palmer 2014).
For household technology choices, insights from the psychology and behavioral economics literatures suggest that both time and risk preferences may help explain the energy efficiency paradox (e.g. Allcott and Mullainathan 2010, Allcott 2011, Gerarden et al. 2015, Ramos et al. 2015, or Schleich et al. 2016). There is some growing body of evidence on the effects of time discounting and risk preferences on energy-efficient technology adoption; however, a comprehensive test of these effects is missing. Researchers have also used different methods to capture time and risk preferences. Some rely on multiple price list experiments, following Coller and Williams (1999) and Holt and Laury (2002), while others have used Likert scales, following Dohmen et al. (2010, 2011) and Falk et al. (2016). While producing internally consistent measures of preferences in a controlled and incentivecompatible manner, experiments are typically lengthy and involve high administrative and financial costs when employed in large sample surveys. In comparison, Dohmen et al. (2011) show that asking individuals for a global assessment of their willingness to take risks is a good predictor of behavior in several domains. While such scale-based measures lack internal consistency and self-reporting is not incentive-compatible, they are simpler, less time-consuming and cheaper than experiment-based measures. Dohmen et al. (2011, p. 543) argue that experiment-based risk measures are good predictors of individual behavior in the financial domain, but are less informative for risk taking in nonfinancial domains. Thus, empirical findings on the role of time and risk preferences for explaining individual behavior and corresponding implications may depend on the type of measure used to elicit these preferences.
Moreover, the effects of loss aversion, that can be expected to affect energy-efficient technology adoption remain largely unstudied. To our knowledge, only Heutel (2017) has – in a parallel effort - empirically investigated the effects of loss aversion on household adoption of EETs. Simultaneously considering the effects of risk aversion, standard time discounting, present bias, and loss aversion on EET adoption to avoids mistakenly conflating their effects. Methodologically, to get internally consistent parameter estimates, the parameters reflecting standard time discounting, risk aversion, loss aversion, and of present bias are calculated jointly at the individual level. Preferences for time discounting, present bias, risk and loss aversion were elicited via (partly incentivized) decontextualized multiple price list (MPL) lotteries. As an alternative, we compare findings when time and risk preferences are elicited via Likert scales. To address concerns with previous literature, adoption decisions were surveyed from decision-makers for low- (LED light bulbs), medium- (appliances) and high- (retrofit) –stake energy-efficient technologies. Further, the study includes many household control variables (such as intention to move, renting, socio-demographics and individual traits), as well as dwelling characteristics such as size or dwelling age. Finally, empirically, our study is the first to utilize representative samples in a cross-country comparison and with a final sample of about 13.500 respondents, our sample size is much larger than in previous studies, allowing for more generalizable results. Thus, our study contributes to the emerging literature that relate preference measures employed in laboratory experiments to actual behavior for representative samples (e.g. Dohmen et al. 2011).
The empirical analysis of EET adoption may also be linked to the conceptual framework of the factors underlying the implicit discount rates (IDRs) developed in BRISKEE deliverable D2.1 and further refined in Schleich et al. (2016). IDRs govern individual adoption of EET in energy-economic models. Since higher IDRs typically imply lower investments in energy efficiency, there is a direct link between EET
(i) preferences such as time preferences, risk preferences, reference-dependent preferences such as loss aversion, and pro-environmental preferences; (ii) predictable (ir)rational behavior, i.e. bounded rationality, rational inattention, and behavioral biases, such as present bias or status quo bias; and
(iii) external barriers to energy efficiency such as split incentives, lack of information, or lack of capital. The empirical analyses employed in this paper include a fairly broad set of these factors, and hence allow for inference on their relation to EET adoption based on a large demographically representative sample. The findings of this report on the factors related to EET adoption also provide the basis for BRISKEE deliverable 2.4., which relates socio-economic factors such as income and age to the factors underlying the IDR, and hence provides guidance for energy-economic modelling. The remainder of this report is organized as follows. Section 2 provides a discussion of the existing literature that links preferences to energy-efficient technology adoption. Section 3 briefly presents the theoretical model of individual preferences, describes the survey and the elicitation of time preferences, risk preferences, loss aversion, and present bias via multiple price lists, and outlines the variables used in the econometric analysis. Section 4 presents and discusses the findings of the econometric analysis. The final section summarizes the main findings and discusses their implications.