Using narratives to infer preferences in understanding the energy efficiency gap

Investing in energy efficiency is crucial for a low-carbon economy, especially in the building sector. Despite ambitious energy targets and various subsidy programs implemented by most developed economies, significant progress is lacking. This "energy efficiency gap" raises questions about whether systematic barriers and household preferences hinder the adoption of cost-effective energy-efficient technologies. This research addresses this gap by using a novel approach: analyzing narratives from homeowners to understand their decision-making processes regarding energy efficiency retrofits. The study focuses on single-family homeowners in the Canton of Zurich, Switzerland, utilizing a large-scale survey that incorporates both open-ended (narrative) and closed-ended questions. The open-ended questions allow for a more nuanced understanding of the reasons behind homeowner decisions, while the closed-ended questions provide a comparative benchmark. The analysis uses NLP to quantify the narratives, providing a scalable and replicable method for studying the energy efficiency gap and technology adoption more broadly.

The paper reviews existing literature on the energy efficiency gap, highlighting the limitations of previous studies. Most past research relies on closed-ended questionnaires, focusing on predetermined explanatory variables. This approach limits the ability to generalize findings and is prone to researcher bias, as it fails to fully capture the decision-makers' thought processes. The study contrasts this with the emerging use of narratives in economics to understand decision drivers. The authors’ method allows for more nuanced insights into the barriers and determinants of technology adoption than traditional closed-ended survey methods. It offers a proof of concept that's easily implemented, scalable to large samples, and replicable across various contexts. The study aims to provide a more robust identification of behavioral barriers and determinants related to energy efficiency, building upon previous taxonomies that categorize barriers into economic, behavioral, and organizational perspectives. The authors propose a refined taxonomy distinguishing between behavioral, financial, non-market, and market barriers.

The research employs a mixed-methods approach, combining quantitative and qualitative data collection and analysis techniques. Data were collected via a survey administered to a large, stratified random sample of single-family homeowners in Zurich, Switzerland. The sample included homeowners whose houses were built before 1990, those who had obtained renovation permits in the preceding five years, and those who had adopted the Minergie energy efficiency certification (this last group was excluded from the main analysis). The survey included both closed-ended (multiple-choice) and open-ended (narrative) questions. The open-ended questions focused on barriers to retrofitting (for non-takers) and determinants of retrofitting (for takers), as well as policy preferences. The closed-ended questions served as a benchmark for comparison. The data analysis involved natural language processing (NLP) to transform narrative responses into quantifiable metrics. A keyword-dictionary approach, involving pre-processing, clustering, and topic extraction, was used to classify the narrative responses into thematic categories. The researchers validated this method through comparison with human coding and analysis of semantic distances using word embeddings. This approach helped overcome some limitations of using solely unsupervised topic models like Latent Dirichlet Allocation (LDA) for short-text responses. Regression analysis was also employed to investigate the correlation between key barriers/determinants and observable household characteristics, building characteristics, policy preferences, and psychographic variables.

The study's key finding is that energy efficiency investments are highly opportunistic. Non-takers (homeowners who did not invest in energy efficiency) often believe their homes are sufficiently energy-efficient, while takers (those who have invested) largely did so due to the necessity of replacing broken or obsolete building components. Financial barriers are substantial, but not the primary driver for takers. Co-benefits – comfort improvements and environmental concerns – play a significant or even more prominent role. The study reveals inconsistencies between responses to open-ended and closed-ended questions, highlighting the value of narrative data in capturing nuanced aspects of decision-making. Table 2 details barriers for non-takers, with the belief that their home is already energy efficient being the most prevalent. Cost was the second most important barrier. Table 3 presents determinants for takers, with replacing broken elements being the most significant determinant. Reducing the environmental footprint and increasing comfort were also important. Table 4 shows policy preferences, with more subsidies being the top suggestion, followed by more information and less bureaucracy. The analysis demonstrates difficulties in targeting policies based on observable household characteristics, indicating a need for more holistic policy design. Regression models (Table 5 and Table 6) exploring heterogeneity in barriers and determinants showed limited predictive power of observable variables, emphasizing the complexity of individual decision-making.

The findings challenge the conventional focus on financial barriers in energy efficiency policies. The opportunistic nature of retrofits, driven by the need to replace broken components, suggests that current subsidy programs may be ineffective and even lead to free-riding. The significant role of non-market benefits such as comfort and environmental concerns underscores the need for policies that incorporate these factors. The observed inconsistencies between open-ended and closed-ended survey responses highlight the limitations of relying solely on traditional stated-preference methods. The difficulty in targeting policies based on observable characteristics points to the importance of understanding the underlying cognitive processes driving individual decisions, emphasizing the need for approaches that elicit more nuanced information.

This study demonstrates the value of using narratives and NLP to understand the energy efficiency gap. The findings highlight the opportunistic nature of energy efficiency investments, the importance of non-market benefits, and the limitations of relying solely on financial incentives. Effective policies should consider institutional factors like bureaucracy and information accessibility in addition to financial incentives. Future research could further explore the use of narratives and AI in understanding complex decision-making processes related to energy efficiency and other sustainability challenges.

The study's findings are based on a survey of homeowners in Zurich, Switzerland, and may not be generalizable to other contexts. The reliance on self-reported data may introduce biases. The NLP approach, while validated, still relies on human interpretation in the topic extraction phase. The study did not specifically investigate the reasons for inconsistencies between open-ended and closed-ended responses, representing a possible area for future research.