The views presented in On Balance are those of the authors and do not represent the views of the Society, its Board, or its members. 

At the 2025 SBCA annual conference in Washington DC, John Whitehead organized a panel session titled “Frontiers in Environmental Valuation” (at Vic Adamowicz’s suggestion). The session was presided over by Tim Haab and the panelists included Mark Dickie, University of Central Florida; Rob Johnston, Clark University; Jonathan Lee, East Carolina University; Frank Lupi, Michigan State University; Kent Messer, University of Delaware; George Parsons, University of Delaware; and Christian Vossler, University of Tennessee. In part the session arose because John and Tim (and Lala Ma) are editing a new Handbook on Environmental Valuation and the session participants were authors on several of the chapters. To kick off the discussion, Tim posed a fascinating question to the panel: “What will environmental valuation look like in 10 years?” The question resulted in a discussion from the panel members as well as several members of the audience. This blog post summarizes some of the themes that emerged, as well as our own thoughts and reflections on the topic. We hope it fosters discussion on the frontiers of valuation and promotes research on valuation.

1. Data

Data for stated preference valuation has traditionally come from surveys. Survey data has changed from mail and telephone surveys to mostly internet panels, although in some cases researchers are reverting to mail surveys or at times in-person surveys because of the sampling properties. But it is increasingly difficult to obtain high quality survey responses. Even official government surveys are finding lower response rates and item non-response. So-called opt-in internet panels (e.g., Dynata) are inexpensive but may be lower quality than more expensive probability-based internal panels (e.g., IPSOS Knowledge Panel). More research is needed here, for example, comparative studies (Sandstrom‐Mistry et al. 2023) and meta-analyses assessing validity across data types. The extent to which opt-in panel data can be used for policy-relevant valuation in the future is dependent upon the results of these future studies.

There are few alternatives to the use of surveys for stated preference researchers – but alternative modes are being explored. One of the more innovative approaches is the use of Large Language Models (LLMs) or synthetic respondents either to test surveys, as complements to surveys, or to substitute for surveys. While there are many concerns about the use of synthetic agents and their ability to reflect actual non-synthetic preferences, there is emerging literature that suggests they can approximate actual choices well. There are also other advantages including low cost, no research ethics requirements (at least at this point), and the potential to obtain information on preferences over time rather than as a snapshot (Collis et al. 2024). Looking ahead, AI and synthetic agents may become powerful complements to traditional surveys, especially for rapid testing of preference structures and longitudinal insights, though direct engagement with human respondents will remain essential for credible valuation.

For revealed preference research a variety of new data sources are being explored. Travel cost and spatial behavior research is exploring the use of cell phone mobility data. Some have experimented with data collection via apps, while others use cell phone “ping” or location data. Databases such as Tripadvisor and Strava are offering anonymized data that describes origins and destinations for large numbers of users. Several papers have used specialized apps / databases like eBird, and it is likely that more such platforms will become available. And sensors, already used for bird and wildlife identification, could be a source of information on attributes and/or visitation. An even more exploratory (and futuristic!) approach is the use of eDNA to identify unique individuals and their locations at sites such as parks (Whitmore et al. 2023). 

For hedonic property value studies there has already been an explosion of data available on housing characteristics and in some countries links to individual information including financing data and migration information. The future could include additional detail on housing attributes that provides insights into variation in environmental features (via sensors or detailed satellite imagery, e.g., Proctor et al. 2025). These emerging data streams also raise important issues of privacy, sampling biases, and governance that the field will need to address alongside methodological innovation.  

2. Endogeneity

Much of the current revealed preference literature struggles with endogeneity challenges (e.g. the challenge of identifying causal relationships between environmental changes and choice behavior from observational data). There was discussion at the session about the need for analysis of more natural experiments (see Wardle 2025). Developing longer term individual panel data sets would help with identification. There is increasing use of application data (e.g., camping permits, hunting permits) that may help identify environmental impacts (Lloyd-Smith and Zawojska 2025). In the future new data sources, including cell phone mobility data, will provide sources of panel observations and help address the various endogeneity issues that arise. These new data sources may help strengthen identification strategies, though many endogeneity challenges will persist given the complex relationship between environmental quality, choice behavior, and social dynamics.

3. Machine Learning and Artificial Intelligence

Several panel members discussed the potential for the use of machine learning (ML) models and AI in valuation. There are several advances in the use of AI in random utility modelling (Sfier et al. 2025; Nova et al. 2025) and researchers are using machine learning models that support causal inference. These will be more common in modelling recreation and property value studies, but also in analyzing stated preference data and benefit transfer (e.g., Johnston and Moeltner 2024). It is also likely that ML techniques will support data fusion – linking the various data sources that are available. But an emerging use of ML is its ability to provide socioeconomic data from satellite imagery and remote sensing data (Proctor et al. 2025). This may be the future of ML in valuation – a way to use various sources of mobile phone, sensor, and satellite data to construct profiles of activities and responses to environmental changes. Continuing integration of ML with welfare economics will require attention to interpretability, transparency, and safeguards that maintain the credibility of valuation estimates.

4. Beliefs and Perceptions

A component that is still lacking in many valuation studies, SP or RP, is the inclusion of beliefs  or perceptions as factors that influence preferences and behavior. Elicitation of beliefs is challenging, and incorporating them into econometric analysis is also difficult. With declining survey response rates and less interest in answering long surveys, how will researchers learn about beliefs? There could be efforts to systematically collect beliefs / perceptions, or it may be possible to use LLMs to approximate beliefs and to use these in valuation. Monitoring of social media may also provide an avenue to develop beliefs profiles. As these approaches develop, maintaining empirical grounding through careful validation against real human beliefs and observed behavior remains important.

5. Strategic Behavior and Hypothetical Bias

The SP literature has witnessed an explosion of research over the past decade on strategic behavior, hypothetical bias, and consequentiality. Beliefs also play a significant role in this literature. Over the next decade we expect continued advances in theoretical and empirical findings that will help in questionnaire design and survey methods to address and detect strategic behavior. There will be a better understanding of the incentives for truthful preference revelation associated with various forms of SP elicitation (single binary choice, repeated binary choice, choice baskets, etc.) with application to public and private goods valuation. These innovations will support more valid and reliable valuation measures.

6. Valuation and National Accounts

Most valuation studies are conducted using a welfare theoretic framework and are aligned with benefit cost analysis or regulatory analysis. But there are increasing demands for non-market values to be included in national income accounts and to provide information to the public and industry on the monetary impact of their activities, including activities that affect environmental quality. This will require a refocusing of analysis and valuation, likely towards markets, exchange values and behavior. Linkages between monetary and physical value will be strengthened. In ten years, we could see a range of different ways that non-market elements of our economy are reflected in regulatory analysis, macroeconomic indicators of wealth, and dashboards of product or industry impacts. Progress in incorporating non-market value into national accounting frameworks is likely to be incremental and may vary substantially across agencies and countries.

7. A Potpourri of Topics

The panelists discussed other emerging topics in the literature including “choice portfolios.” Most environmental valuation studies focus on relatively simple choice sets (e.g., recreation alternatives, housing alternatives, products with environmental certification, etc.). But consumers may consider packages or sets of alternatives. These portfolios may be better reflections of behavior and thus provide more accurate valuation measures. There was also discussion of themes that have been present for years but continue to evolve or be integrated into valuation. These include how uncertainty is dealt with in valuation research and how behavioral economics is integrated into valuation research. There is an interaction between behavioral economics and behavior under uncertainty including topics like the use of heuristics (e.g., attribute non-attendance), deviations between perceived and objective risks, and other topics. In the future we expect better behavioral models that reflect improved knowledge about behavioral responses to risk, perceptions, networks, etc. In addition, several panelists raised how uncertainty, expectations, and learning under climate change create new opportunities for modeling behavior in dynamic environments.

8. Global Challenges

The use of environmental valuation in regulatory and policy analysis has increased over time, particularly values related to human health and the environment. But environmental valuation still plays a relatively small role in the quantitative analysis of some major global challenges. Ten years from now environmental goods and services will be included in the quantification of impacts and development of policy responses to climate change (including assessment of extreme events, adaptation to change, learning, and incorporating expectations) and biodiversity loss (through translation to recreation / tourism values, property values, and improved passive use valuation). Several panelists emphasized that future applications must also attend to distributional consequences and equity considerations, particularly when addressing risks affecting vulnerable communities. One can imagine incorporation of non-human preferences (extensions of animal welfare) and altruistic preferences through improved theory and methods.

Looking Ahead

Environmental valuation is entering a period defined by richer data, faster computation, and deeper integration with policy needs. The opportunities are exciting. New tools such as synthetic agents, mobility and sensor data, and machine learning can help illuminate behavior and value environmental change in ways not previously possible. Yet the discipline’s continued impact will depend on maintaining credibility, grounding innovations in real human preferences, and communicating results clearly to decision makers and the public. Progress across these fronts will ensure that valuation remains both rigorous and relevant as society confronts climate change, biodiversity loss, and other critical environmental challenges in the decade ahead.

The Panel:

• Organizer: John Whitehead (Appalachian State University)
• Presiding: Tim Haab (The Ohio State University),
• Christian Vossler (University of Tennessee)
• Mark Dickie (University of Central Florida)
• Jonathan Lee (East Carolina University)
• Rob Johnston (Clark University)
• Frank Lupi (Michigan State University)
• Kent Messer (University of Delaware)
• George Parsons (University of Delaware)