Discrete choice modeling in environmental and energy decision-making: an introduction to the special issue

Abstract

This editorial introduces the Special Issue on the role and use of discrete choice modelling in informing environmental and energy decision making and summarises the main insights provided. The papers collected in this Special Issue address a range of applications of discrete choice modelling, from energy decisions to climate change and transport choices, from tourism preferences for climate change mitigation to wildlife conservation strategies and ecosystem management. However, all the papers share the same fundamental aim of using novel methodological approaches in choice models to tackle conceptually and spatially specific policy questions and support decisions with robust empirical evidence. This is particularly relevant to enhance the real-world policy uptake of choice modelling results and develop a network of practice and capacity building to improve the quality of information that is provided to stakeholders and decision makers.

Keywords:

• Choice modelling
• environmental decisions
• energy decisions
• policy making

Decision-making is the process of supporting multiple policies or courses of action and can be a complex and fast-paced stage of the policy cycle. Policy decision-making in energy and environment involves interdisciplinary and cross-sectoral choices among the multitude of strategies available to conserve energy, reduce pollutant emissions, mitigate and adapt to climate change, and improve the management of natural capital. Decision-making for the transition to a sustainable society needs to tackle diverse challenges, among which individual and organizational behaviors are at the core. Therefore, environmental and energy policy making can be improved through a better understanding of individual choices in different contexts, time and space horizons, and uncertainties resulting from human behavior, technological development, and social systems.

The set of tools and data available to gauge this improved understanding of choices is rapidly expanding and discrete choice models (DCMs) represent a promising methodology to investigate the effects of policy decision making. DCM is the wide family of stated preference techniques (e.g. choice experiment, conjoint analysis and contingent valuation) applied in environmental and energy economics, transportation, health economics and marketing since the early 1980s. DCM are extensively employed in many fields directly or indirectly linked with environmental and energy decisions, with hundreds of applications ranging from emissions reduction policies (Williams and Rolfe 2017), to plastic litter in the marine environment (Abate et al. 2020), to groundwater released from mining activities (Legg et al. 2020). The popularity of environmental and energy applications of DCM lies in the range and flexibility of information provided on individual and aggregate choice behavior, which is a critical element for informed and effective policy making. This, in turn, improves policy decisions in a variety of contexts, such as the development and evaluation of public policies, the allocation of resources, and the design of markets for energy and environmental services.

Haghani, Bliemer, and Hensher (2021) conducted a 50-year DCM literature review and report a steady growth of published papers focusing on the value of travel time, willingness to pay, and consumers’ preferences. These measures are frequently used in policy decision making for supporting investment in innovative transport systems where the travel time saved is the policy objective, or in environmental and public policies where the willingness to pay or consumers’ preferences are included in cost-benefit analysis or impact assessment analysis. DCMs provide information that can be used to: investigate the preferences for characteristics or attributes of a specific policy prior to implementation; predict the responsiveness to policy-related changes in environmental and energy goods and services; and produce a set of synthesis measures for direct policy decision use. Welfare measures expressed as the average marginal willingness to pay or value of travel time saved can be used directly in the context of cost-benefit analysis for the valuation of transportation or environmental initiatives. DCMs provide several advantages within a cost-benefit analysis context to elicit monetary values flexibly and robustly for non-market externalities. National and international policy appraisal and cost-benefit guidance emphasize this crucial policy use of DCM outputs (HM Treasury 2020; OECD 2018). Louviere, Hensher, and Swait (2000) report in their book diverse applications of DCM where the method was pivotal to determining the users’ demand and as input into a cost-benefit analysis, from high-speed rail and toll road investments to forestry impacts on recreation and wildlife. The authors report how DCM can facilitate the estimation of demand directly and cross-elasticity effects, individual preferences, and willingness to pay. These economic indicators can also be relevant for segmenting respondents into clustered groups or to perform further analysis on conditional posterior distributions such as spatial models (Budziński et al. 2018).

However, Haghani, Bliemer, and Hensher (2021) do not report whether the reviewed studies are adopted in actual policy decision making. While it is recognized that DCM are often performed to directly or indirectly respond to a policy need (Louviere, Hensher, and Swait 2000) and the policy relevance of DCM outputs is widely acknowledged, a closer integration between methodological advancements and real-world policy uptake is still needed (Chandoevwit and Wasi 2020) and, in reality, only a subset of DCM are in fact designed for a specific policy uptake. Moreover, during recent decades, DCM applications have greatly increased in complexity, from the use of machine learning techniques (Van Cranenburgh et al. 2022) to application in emerging fields such as moral decisions and social sciences (Liebe and Meyerhoff 2021) to the inclusion of non-observable and latent constructs (Mariel and Meyerhoff 2016). Increased complexity of modeling strategies and methodological advancements have not always translated into a more informative use of DCM outputs in policy-relevant decision making (Chorus and Kroesen 2014).

In this Special Issue, we present a portfolio of DCM applications in the field of energy, climate change and transport choices, climate change and tourism preferences, wildlife conservation strategies and ecosystem management. The main objective of this Special Issue is to discuss the gap between DCM applications and policy decision-making through papers proposing innovative choice modeling developments and empirical analyses to address real-world environmental and energy management policy issues. A shared feature of the papers is to tackle conceptually and spatially specific policy questions by using novel methodological approaches in choice models to support decisions with robust empirical evidence.

DCM use in environmental and energy decision-making can be prospective (before the policy decision is made) or retrospective (once the decision is applied). Prospective analysis focuses on potential new products, policies and programmes and retrospective analysis focuses on observed behavior to investigate consumers/users’ preferences. Grilli et al. (2021), Notaro and Grilli (2021), Waygood et al. (2021), Pelletier et al. (2021) and Raffaelli et al. (2021) in this Special Issue represent examples of prospective analysis, whereas Zong, Zhang, and Jiang (2021) is a retrospective analysis. In prospective analysis, the design of the DCM survey is the most crucial and controversial aspect and multiple solutions have been proposed to mitigate cognitive biases to produce valid and reliable estimates (Johnston et al. 2017). Possible solutions can be applied in designing the questionnaire (ex-ante solution) or/and during the econometric analysis (ex-post). Raffaelli et al. (2021), Waygood et al. (2021) and Pelletier et al. (2021) propose innovative strategies to improve DCM questionnaires and mitigate hypothetical bias, capture respondents’ latent traits and investigate the role of information in forming the final choices. Strengthening the practice of designing DCM questionnaires is crucial for fostering robust results. At the same time, in both prospective and retrospective analysis, the modeling of choice responses is also fundamental to produce informative DCM outputs. DCM responses are traditionally analyzed with multinomial logit models, but since the paper by McFadden and Train (2000), data modeling has improved. Articulated mixed models and hybrid models represent the latest proposed and tested approaches (Hensher and Johnson 2018) which can provide a richer interpretation of responses.

Zong, Zhang, and Jiang (2021) contributes the only retrospective paper of this Special Issue presenting on the energy sector. The authors demonstrate how the Hierarchical Archimedean Copula, which is a copula-based multivariate ordered probit model, can improve the categorization of energy consumers in in-home energy consumption (electricity and gas, i.e. the domestic sector) and out-of-home energy consumption (gasoline, i.e. the transport sector). The paper contributes to understanding the heterogeneity of energy users with the aim of improving management in the energy sector by using results to propose tailored incentives for private and commercial customers or to inform and fine-tune national level energy strategies and policies.

Waygood et al. (2021) and Raffaelli et al. (2021) contribute to the climate change literature by investigating the role of CO₂ emissions in car purchase decisions and tourists’ preferences for decarbonizing strategies, respectively. Waygood et al. (2021) demonstrate how DCMs can help in our understanding of how transport preferences are intertwined with attitudes and beliefs. They jointly study the car purchase choices and pro-climate attitudes and individuals’ behavior. The authors compare the well-known New Environmental Paradigm and General Ecological Behavior scales with a simple Climate Change-State of Change question to conclude that results are comparable, although research is needed to characterize respondents’ attitudes and behaviors for tackling climate change. The role of framing of CO₂ information related to car choices is also investigated to conclude that framing reduction targets in terms of tonnes per year and from a society’s reduction target point of view, is better perceived. The suitability to use a simpler Climate Change-State of Change approach for environmental motivations could translate into a more direct use of such measures in decision-making, through diminished information and modeling burden for practitioners and a clearer synthesis indicator for policy makers. Results could also be used in setting new framing and communication strategies for emissions in the transport sector.

To investigate the most successful decarbonizing strategies in tourist facilities in the north of Italy, Raffaelli et al. (2021) present a choice experiment which dedicates special attention to ex-ante strategies to mitigate hypothetical bias. A split sample survey was used to obtain individual and collective preferences for tourists of the mountain areas. Results reveal that hypothetical bias inflates WTP estimates, and mitigation strategies are needed to produce more refined welfare measures relevant for policy decision makers. The attitude of respondents toward net-zero tourist facilities is positive but their willingness to pay is still very low to motivate private decarbonization efforts. Results suggest that effective policy decisions in this field would include joint public and private initiatives to decarbonize the sector and awareness campaigns to accelerate climate neutral tourism.

Pelletier et al. (2021) investigate the role of information provided to respondents in choice experiments. This is a widely debated issue in choice experiment development and there is still uncertainty around the optimal level of information in case studies concerning complex ecological processes. In a split sample survey on improvements to a wetland habitat in the east of Australia, the authors investigate whether different levels and types of information provided to respondents influence perceptions of the survey and the preference heterogeneity for some of the choice experiment attributes. Findings suggest that background information has limited impact on heterogeneity and elicited willingness to pay values. In line with Waygood et al. (2021), results in Pelletier et al. (2021) reinforce the possibility of reducing the burden for practitioners and policy makers, as less information may be needed, in line with cost considerations.

Findings in Pelletier et al. (2021) also highlight the direct policy relevance of DCMs in ecosystem management. Taking a similar ecosystem as a case study, Grilli et al. (2021) examine the potential and limitations of using DCMs in developing local level ecosystem accounts for saltmarsh areas in the East of England. The authors argue that modeling choice data to provide simulated exchange values that are in line with ecosystem accounting guidelines (SEEA EA 2021) can result in a more informative and robust set of evidence for environmental planning and policy purposes. Results highlight the flexibility of DCM use in environmental decision-making. Local land-use planning could benefit from better and more comprehensive information to inform restoration and conservation. Moreover, estimates that are comparable with exchange values can help to better assess tradeoffs between environmental policy and socio-economic outcomes.

Programmes aimed at protecting large carnivores are a controversial issue in biodiversity policy making. Notaro and Grilli (2021) study the willingness to contribute to wildlife management programmes in the Italian Alps. The authors use a DCM to derive preferences for conservation strategies and show that the protection of wolves and lynx is a welfare-increasing policy. Results are of particular importance in showing the role of DCMs in unveiling benefits of environmental policies which would otherwise be overlooked since this information is difficult to measure. Findings can, therefore, be directly used in the appraisal of public policies to manage wildlife and as a benchmark to develop financial instruments such as fees and entrance tickets to natural areas which would also positively impact the tolerance of local communities toward human-wildlife coexistence.

In conclusion, all the DCM applications presented in this Special Issue have clear implications for environmental and energy decision-making. However, only a subset of these studies were actually designed to directly inform policy decision-making. The Grilli et al. (2021) study was explicitly designed within the Marine Pioneer programme, a UK-wide pilot project funded and supported by the Department for Environment, Food and Rural Affairs and aimed at informing the delivery of the Government’s 25 Years Environmental Plan (DEFRA 2018). Findings reported in the Grilli et al. paper were also discussed with, and presented to, local stakeholders and associations promoting the preservation of the saltmarsh area in the Deben Estuary, in eastern England. In the Waygood et al. (2021) study, authors benefited from the support of several policy actors, including the Minister of the Economy and Innovation, the Minister of the Environment and Climate Change, Transport Canada, and Natural Resources Canada. The research was supported through funds put aside by the Government for electrification and climate change and the results were used to inform decision-makers about how the current new vehicle labeling for climate emissions can be improved. Similarly, the Pelletier et al. (2021) study was part of a multi-faceted project on flooding prevention in the Richmond River catchment area, Australia. The project involved close engagement with local stakeholders and agencies and the DCM supported the need to estimate non-market values that would aid in channeling investments to reduce the impact of flooding events.

Overall, while half of the studies presented in this Special Issue were directly aimed at informing real-world policy decisions, a comprehensive assessment of how and at what stage of the policy cycle the research findings have been used, does not exist. This is the case for most of the DCM outputs. The lack of evidence and assessment on the use of DCM results in policy decisions does, in fact, limit the knowledge transfer between academia and decision makers in terms of policy demand, actual use and DCM capacity building.

Acknowledgements

The authors would like to thank the organisers and participants of the International Choice Modelling Conference 2019 held in Kobe, Japan. This Special Issue is a collection of the papers presented during the conference session “Energy and environmental decision-making.”

Disclosure statement

No potential conflict of interest was reported by the authors.