Results from a survey of life cycle assessment-aligned socioenvironmental priorities in US and Australian communities hosting oil, natural gas, coal, and solar thermal energy production

Large energy infrastructure is often socially and environmentally disruptive, even as it provides services that people have come to depend on. Residents of areas affected by energy development often note both negative and positive impacts. This reflects the multicategory nature of socioenvironmental outcomes and emphasizes the importance of careful, community-oriented decision making about major infrastructural transitions for processes like decarbonization. Quantitative tools like life cycle assessment (LCA) seek to collect and report comprehensive impact data, but even when successful, their value for decision support is limited by a lack of mechanisms to systematically engage with values-driven tradeoffs across noncommensurable categories. Sensitivity analyses designed to help decision makers and interested parties make sense of data are common in LCA and similar tools, but values are rarely explicitly addressed. This lack of attention to values—arguably the most meaningful set of decision inputs in such tools—can lead to overreliance on single issue (e.g. climate change impact) or proxy (e.g. monetized cost) outputs that reduce the value of holistic evaluations. This research presents results from preregistered hypotheses for a survey of residents of energy-producing communities in the United States (US) and Australia, with the goal of with the goal of uncovering energy transition-relevant priorities by collecting empirical, quantitative data on people’s priorities for outcomes aligned with LCA. The survey was designed to identify diverse value systems, with the goal of making it easier for users to identify and consider value conflicts, potentially highlighting needs for further data collection, system redesign, or additional engagement. Notably, results reveal remarkably consistent priority patterns across communities and subgroups, suggesting that the common LCA practice of equal prioritization might be masking decision-relevant information. Although this effort was designed specifically to support research on energy transitions, future work could easily be extended more broadly.


Introduction
The fossil fuel-dominated energy system must transition to zero-carbon resources or continue to exacerbate climate change, a major threat multiplier and likely ongoing contributor to global inequality and injustice (Intergovernmental Panel on Climate Change 2014, Schlosberg and Collins 2014). Such a transition comprises two major processes: deindustrialization in the form of closing existing fossil fuel infrastructure, and industrialization in the form of dramatically expanding the zero-carbon energy system (Grubert 2020). Both processes have major implications both for society overall and for communities hosting new and retiring infrastructure.
The timeline of transition is fast in the context of industrial systems, but relatively slow from the perspective of human experience. Meeting Paris Agreement targets of constraining global warming to well below 2 degrees Celsius, and preferably to 1.5 degrees Celsius, suggests that global greenhouse gas (GHG) emissions need to reach net-zero by mid-century. In the United States (US), the Biden administration has called for decarbonization of the electricity sector by 2035, and of the full economy by 2050 (White House 2021). As a result, the major infrastructural transitions necessary to decarbonize while responding to climate change will take place over a few decades-shorter than the lifespan of much of the relevant infrastructure, but long enough to require sustained commitment through challenging conditions on the part of the people affected (Grubert and Hastings-Simon 2022). This goal is also expected to lead to mass electrification of traditionally fossil fuel-based applications like transportation and home heating to enable use of mature zero-carbon systems like wind, solar, and geothermal plants, which in turn means a substantial expansion of the electricity sector (Williams et al 2021) and a noticeable change in the energy infrastructures people directly interact with. As an example, the US can expect to install tens of thousands of new zero-carbon power generators, while simultaneously closing about 10 000 fossil fuel-fired generators that currently provide most of the country's electricity; and either downsizing (and ultimately retiring) or fundamentally repurposing fossil fuel supply infrastructures like mines, wells, and pipelines, while simultaneously retiring people's internal combustion engine personal vehicles, natural gas stoves, and natural gas furnaces. Given global fossil fuel dominance and the urgency of transition, the picture is similar around the world, with varying considerations about early retirements, costs, and distributional justice associated with the transition (Pfeiffer et al 2016, Tong et al 2019, Cui et al 2021. One of the core challenges of the normative decarbonization transition is managing change in a way that promotes justice, including by correcting historical harms (Bullard 1999, Evans and Phelan 2016, Reames 2016, Malin et al 2019, Batres et al 2021, Bozeman et al 2022, while minimizing environmental harms beyond GHG emissions alone (Tarroja et al 2020). The pursuit of 'sustainability' that includes both justice and direct environmental objectives through a suite of social, environmental, and financial criteria is necessarily subject to tradeoffs that arise from pursuing multiple and sometimes conflicting goals across issue areas. Multicriteria decision support tools like life cycle assessment (LCA) are designed to provide rigorous information about systems across these multiple criteria in order to inform decisions about those tradeoffs. Ultimately, though, even when experts are able to generate and communicate extremely high quality information about what impacts might occur, how severe they might be, and how certain they are, tradeoffs are a matter of values, used here in the sense of human preferences (Brown 1984). Not all values are easily aligned with the types of analysis that are granted authority in decision processes (or even quantifiable) (Chan et al 2016, Grubert 2018a, Rawluk et al 2019, which reflects a fundamental limitation of decision support tools and requires humility from analysts. Nonetheless, within the context of decision support tools, the ability to quantitatively and rigorously incorporate values as an organized set of magnitudes of preference (Brown 1984, Tadaki et al 2017, Zuluaga et al 2021 aligned with the issues that have been defined as relevant to the decision context (ISO 2006) can extend the usefulness of these tools for decision makers tasked with weighing challenging tradeoffs with societal implications.
The need to explicitly account for values in LCA and related methods has long been recognized (Heijungs 1994, Finnveden 1997, Hofstetter et al 2000, Freidberg 2015, Rosenfeld and Ptolemy 2017. Prior work has described efforts to incorporate values in forms like preference-based weighting factors (Mettier et al 2006, Grubert 2017c, Ekener et al 2018, Tarne et al 2018. One ongoing challenge is that empirical data on preferences across impact categories used in LCA (including social, environmental, and financial issues) is difficult and often costly to generate rigorously. Data tied to specific decision contexts can be difficult to generalize, especially given a lack of high resolution data across diverse contexts. Even without the ability to fully generalize, however, more and more diverse data aligned with analytical methods (rather than specific projects) can inform understanding of archetypical value patterns. For example, understanding that people consistently and repeatedly prioritize certain issues over others can inform approaches to data collection, presentation, and other issues.
To contribute to this literature, this paper presents results from preregistered analyses of survey data on preferences across LCA-aligned socioenvironmental outcomes, collected in 2015 and 2016 from residents of US and Australian communities hosting oil, natural gas, coal, and solar thermal energy production. These empirical data contribute to the literature on values in LCA, as well as to the literature on community attitudes about energy transition. Although the data are not designed to be representative or fully generalizable, the focus on energy infrastructure host communities (but not on specific projects) means that these data reflect the preferences of people who have in many cases experienced many of the types of impacts that LCA seeks to measure, and that are expected to continue as the energy transition proceeds. By investigating such preferences at high topical resolution (30 impact categories) using a consistent method in numerous host sites (ten regions, including multiple communities in some cases), this work contributes an unusual empirical dataset on preference that can inform not only LCA practice, but also broader work on community consultation, policy development, and impact assessment related to energy transitions. Many of the people who participated in this study live in relatively isolated communities and experienced specific types of disruptions (due to the timing of the study), and the use of approaches like census tract-level saturation mailings for data collection enabled access to their perspectives in a way that more common survey approaches (e.g. web panels) cannot.

Background
Both fossil industry contraction and zero carbon industry expansion have wide-ranging implications for society as a whole, and particularly for the specific communities hosting these changes (Haggerty et al 2018). Closing infrastructure means losing or transitioning jobs and tax bases (Newell andRaimi 2018, Pai et al 2020), sometimes accompanied by challenges to social identity (Colvin et al 2015). Where new facilities are developed, host communities might gain jobs and taxes. Simultaneously, industrialization, land use change, construction boom dynamics, and other processes can be major challenges regardless of resource type (Devine-Wright 2009, Colvin 2020, Bessette and Mills 2021. Impacts might be inequitably distributed and geographically concentrated, contributing to the potential for physical, economic, and cultural dislocation (Bluestone 1982, Smith 2019) that can contribute to strong political coalition formation during transition. Meanwhile, active intervention in transition by carbon-based industry actors (Mildenberger 2020, Franta 2021) exacerbates incumbency bias and threatens the success of a normative decarbonization transition (Grubert and Hastings-Simon 2022).
Processes with substantial societal benefits (e.g. decarbonization and accompanying impact reductions for air pollution, water consumption, solid waste generation, and other issues related to fossil fuel consumption) can still have real and negative local impacts (Haggerty et al 2018, Raimi et al 2022. One person's local benefit might be another's local harm (Graff et al 2018). Community consultation processes have long been a point of tension for rapid industrial shift, as local and broad societal goals might be at odds (Trigger et al 2014, Walsh et al 2017, Owen et al 2022, Sovacool et al 2022. Even within a given geography, impact disparities can be substantial (Giang and Castellani 2020). Either a community has a meaningful ability to consent (i.e. because veto is possible), or not, and individual groups or people within a community might hold dramatically different views of preferred futures (Grubert and Skinner 2017).
When a community's consent is withheld from projects with broad support outside the community, e.g. renewable energy projects, accusations of NIMBYism are common and frequently based on a relatively coarse understanding of potential drivers of opposition that might mischaracterize opposition as poorly informed or unreasonable. In fact, just because a type of project can be well implemented does not mean that a specific project has been (Burningham et al 2006, Devine-Wright 2009, Konisky et al 2020, Measham et al 2021, Sward et al 2021. Simultaneously, the scale of impact from ongoing fossil fuel use (including climate change) is so severe, and so urgent, that delays related to local impacts that might be perceived-rightly or wrongly-as relatively tolerable (e.g. noise, traffic) can lead to resentment outside the community that tends to reinforce otherization. Recognizing the social context and political nature of a just transition that accounts for distributional impacts, material well-being, and power dynamics, and prioritizing earning and retaining trust, is core to successful decarbonization processes (Miller et al 2015, Healy andBarry 2017).
Public attitude elicitation related to energy systems transformation frequently focuses on attitudes toward specific projects and processes (e.g. (Boudet and Ortolano 2010, Jacquet 2012, Olson-Hazboun et al 2016, Junod and Jacquet 2019, Tan et al 2020). Although high resolution information is crucial when evaluating specific developments, it might not translate well to more abstracted contexts, like policy design (Batel and Devine-Wright 2015). In such contexts, understanding tensions across value systems that are normative, aspirational, and generalized is important well before proposals are sufficiently concrete to engage specifically. For example, attitudinal elicitation focused on a specific project might reflect community members' understanding that particular types of information are likely to be privileged in decision making. This recognition possibly leads to stated emphasis of less-important priorities due to their perceived power in the process overall (Fischhoff 2013). Particularly in contexts where experts define the types of impacts that are in scope for a decision (e.g. environmental impact statements, LCA, and others), understanding how people with diverse backgrounds value those impacts in general rather than in a specific case can be especially valuable for grounding discussion. Highly specific information can be difficult to translate to more categorical preferences for different types of socioenvironmental outcomes due to emphasis on particular local infrastructure needs or tensions. For example, a group might value air quality over additional jobs in general, but advocate specifically for attention to job creation if air quality is already high but recent layoffs caused concern-a contextual preference that can mask the absolute preference for excellent air quality. In a policy or early stage design context, highlighting how a particular action might affect highly valued outcomes can inform targeted attention to burden shifting from addressing deficits in other areas. Put another way, assessing both contextual and aspirational priorities can inform decisions that need to balance urgency and importance.
In practice, quantitative decision support tools focused on sustainability deal poorly with values, whether local or national, contextual or aspirational. This weakness is partially due to an interest in presenting 'objective' data, although the nature of multicriteria sustainability assessment precludes a single correct answer as to what is most 'sustainable' and necessarily includes numerous value choices (Grubert 2017c).
The field of decision analysis shows that making subjective decision inputs like risk attitudes and preferences explicit can improve decision making, and decisions can be normatively better without being objective (Howard 1988). When sustainability is understood as a multicriteria concept, combining measures of impact across noncommensurable categories (e.g. community future, GHG emissions) in a way that facilitates active and informed engagement is one of the core challenges for decision support. This challenge particularly emerges because it depends on acknowledging that conflicting perspectives can nonetheless be equally valid, requiring space for consideration of multiple truths (Ekener et al 2018). Multicriteria sustainability assessment tools like LCA attempt to quantitatively and simultaneously evaluate all relevant impacts associated with the object of analysis (the functional unit) in a single framework, but suffer from challenges associated with effectively identifying and contextualizing competing value systems that enter into inter-and intra-category normalization and weighting for communicating results. For example: how are weighting factors derived, who is involved, and how are conflicts evaluated (Heijungs et al 2010)?
There is no formal intercategory prioritization schema in typical LCA practice. In fact, weighting across categories is actively proscribed by international standards for comparative results being presented to the public, though not all LCA follows these standards (ISO 2006). Although imposing a single value-based prioritization weighting on results has high potential for manipulation, it is also true that prioritization is inherent to multicriteria evaluations because of the need to convert information about more than one issue into a single decision or recommendation (Grubert 2017b). As such, approaches that allow for consideration of multiple and diverse prioritizations of different categories of impact could have particular value both for evaluating decision robustness (e.g. would I still choose the same outcome if I cared deeply about air pollution versus water consumption?) and for identifying points of conflict that could motivate additional data collection or system redesign (e.g. if I would make an extremely different decision while prioritizing air pollution over water consumption, do I need higher resolution data on those impacts to confirm they are real? Could I redesign the alternatives to reduce the impact disparities?).
Current practice in sustainability assessment generally either implicitly or explicitly weights different categories of impact as equally important (Pelletier et al 2019), which can mask clear and measurable preferences that people hold. Preferences, like prospects and risk attitudes, are fundamental inputs to rigorous decision analysis (Howard 1988). Particularly because LCA results are highly sensitive to value choices that might not be clearly grounded (De Schryver et al 2011, Prado et al 2019, the masking of preferences is especially problematic if people generally agree about which issues are more or less important (Grubert 2017c). For example, an alternative that performs especially well on low-priority issues but poorly on high-priority issues might appear to be as 'sustainable' as one with the opposite profile in a model, but not to the host community. As such, developing transparent, replicable methods for evaluating alternative prioritization schemas is a high priority for improving the performance of tools like LCA for decision support.
Evaluating preferences across multiple sustainability criteria in a manner that can be directly integrated to LCA could potentially elevate its value as a decision support tool (Grubert 2017c). Despite interest, LCA has been less effective for public policy decision support than it could be, in part due to process concerns like a lack of guidance on how to compare across noncommensurable impact categories and a failure to integrate stakeholders throughout the process (Seidel 2016). Multicriteria decisions commonly require tradeoffs and compromises, which, in turn, require the application of value judgments like preferences in combination with physical data like quantities (Dietz 2013). Making such value judgments explicit, and designing systematic ways to integrate multiple and potentially conflicting value judgments to sustainability assessment in a manner that enables public access to, participation in, and satisfaction with decision processes (Endter-Wada et al 1998, Lazarevic 2015, Sangaramoorthy et al 2016, can align LCA with best practices from decision science. As with other elements of LCA practice, such as intra-category normalization metrics called characterization factors (e.g. global warming potentials that are used to express the relative radiative forcing associated with different GHGs) (Bare 2002), ensuring that prioritization schemas are rigorous, scientific, and explicit is critical. Because value systems are not fixed or singular, using multiple value profiles to evaluate different scenarios during decision making-rather than attempting to calculate a singular result-is advisable and emphasizes again that these tools are decision support, not decision making, tools. Where values conflict (Fleurbaey and Zuber 2013), understanding the nature of those values and the conflict drivers can be informative for identifying points of conflict that could be good candidates for further information gathering, consultation processes, or other follow-on activities (Grubert 2017c).
In part because of the large burden associated with eliciting preferences, both on communities (Seidel 2016, Walsh et al 2020, Taylor et al 2021 and researchers (Grubert 2017a), data on prioritization across socioenvironmental categories that are not explicitly related to a specific decision process are rare. Much research on attitudes addresses either persuasion (e.g. 'social acceptance') rather than measurement (Pooley andO'Connor 2000, Raimi andLeary 2014) or highly context-specific information that might not be well suited to informing more general evaluations of socioenvironmental impact. This contextual challenge particularly arises when researchers' motivations are believed to be in line with a viewpoint that leads to systematic nonparticipation by people with specific attitudes (de Rijke 2013, Friedl and Reichl 2016). As such, efforts to include preference data in LCA-particularly tabletop LCA that might not include active engagement with decision makers-is challenging in part due to a lack of data. Having basic, LCA-aligned information about patterns of preferences across impact categories (for all forms of LCA) could be particularly useful in evaluating potential points of contention across divergent long-term scenarios for processes like energy system design ahead of specific project decisions (e.g. (Williams et al 2021)). Generalization from specific to aspirational priorities should be performed cautiously and might not be appropriate for evaluating specific project decisions, but could be used for 'hot spot' analysis (as with some forms of social LCA (Benoit-Norris et al 2012)) and might improve as additional empirical data are collected.
LCA could benefit from access to quantitative data that allows for values-based weighting to support analysis (Ekener et al 2018). Data reflecting empirical patterns of priorities across impact categories used or proposed for use with social and environmental LCA, and for life cycle sustainability assessment that combines these with life cycle costing (Finkbeiner et (Grubert 2017c). That method is effectively an additional impact assessment method that tests sensitivity to empirical societal preferences across impact categories. Practically, this approach takes vectors of preference factors as inputs for intercategory weighting (analogous to intracategory characterization factors like the Global Warming Potential, or GWP). Different vectors represent different patterns of preference, selected for diversity rather than representativeness in order to evaluate the robustness of interpretive conclusions to diverse but empirical value systems expressed in the form of preference weighting factors across impact categories.
This work is part of a broader mixed methods project on socioenvironmental attitudes in energy host communities, with methodological details (Grubert 2017a(Grubert , 2017c(Grubert , 2019 and other results (Drummond and Grubert 2017, Grubert and Algee-Hewitt 2017, Grubert and Skinner 2017, Grubert 2018b) described elsewhere. The survey was conducted in communities hosting selected types of energy production (oil, natural gas, coal, and solar thermal) in the US and Australia, without explicit reference to or emphasis on specific ongoing projects. The goal was to assess LCA-aligned preference patterns in geographic communities that have experience with a wide range of energy-related socioenvironmental impacts, rather than to comprehensively evaluate experiences of energy infrastructure and energy transition. The lack of anchoring to a specific project, in combination with the use of the same questionnaire at multiple sites hosting different kinds of energy infrastructure, means that this dataset offers opportunities not only for integrating empirical, descriptive attitudinal data (Dietz et al 2005) to LCA, but also for directly comparing preference patterns across sites. Beyond the application for LCA, this approach contributes to the broader literature on both general and specific attitudes at multiple scales (Batel andDevine-Wright 2015, Demski et al 2015) by investigating overall value systems contextualized by place, with places selected because of their proximity to energy development, rather than attitudes about specific or general energy projects. The remainder of this paper describes the survey and associated preregistered hypotheses, then presents and discusses results.

Methods
The theoretical grounding, design (including copies of the questionnaire and other materials), fielding strategy, and response rate for the survey described in this paper are extensively reported elsewhere (Grubert and Skinner 2017, Grubert 2017a. The work was performed under Stanford University Institutional Review Board Protocol IRB-33232. In brief, a three-contact mail survey of unnamed persons was fielded in the US (12 333 households) and Australia (1000 households) in 2015 and 2016, distributed across ten regions, primarily using saturation mailing at the postal carrier route level. The instrument was designed to critically investigate LCA-aligned socioenvironmental priorities of energy host community residents in a place-based context, using approaches like forced tradeoffs and emotionally-grounded questions (e.g. asking participants to qualitatively describe their most and least favorite things about their natural environment and local community before proceeding to quantitative ranking and rating tasks) (Sherren et al 2021).
Locations were selected because of their role as energy infrastructure host communities, focused on oil extraction (Eagle Ford-Texas, US; Bakken-North Dakota and Montana, US; Oklahoma, US); natural gas extraction (New South Wales, Australia); coal extraction (Powder River Basin-Wyoming, US; Bowen Basin-Queensland, Australia); and solar thermal electricity generation, with communities selected based on proximity to specific solar thermal projects (Crescent Dunes-Nevada, US; Ivanpah-California, US; Genesis-California, US; Desert Sunlight-California, US). Mail surveys (with a link to a web alternative on the third contact postcard) were selected in part due to the infeasibility of web surveys in many of the selected regions, many of which included very small communities (Grubert 2019). Very low returns (five responses) from a community in Queensland mean this location was not analyzed separately. An additional region (urban Pennsylvania, with distributions in Pittsburgh and Philadelphia) was surveyed using the same questionnaire by web, in part due to better access and observed challenges with mail surveys in denser regions. Overall, the intent of the work was to identify diverse empirical preference patterns using a consistent instrument rather than to conduct a random sample of representative sites for statistical extrapolation. Similarly, as the intent was to identify stated values with low salience and limited concretization, the survey itself was designed with low salience. Grubert (2019) reports and contextualizes response rate (overall, 5%-6%, AAPOR RR1; 14%-17%, AAPOR RR3). In total, 1311 responses were included in this analysis (table 1).
Mail survey data were entered into a spreadsheet by hand, and web survey data were imported from Qualtrics. This paper analyzes only quantitative survey responses and demographic data, although the survey also included several open-ended questions for short qualitative responses. The survey questions considered in this analysis are included verbatim in table 2, with question ID listed to improve interpretability of the supplementary data file.
Responses to question B1 were given in the form of rankings (ten categories, mixed social and environmental), while rankings were constructed from group rating means for questions B2 (15 categories introduced as environmental considerations) and B3 (15 categories introduced as social considerations). Ranking questions were posed as unipolar five-point items from 'not at all important' (coded 1) to 'extremely important' (coded 5).
Quantitative responses were unaltered, except to normalize quantitative responses about a hypothetical budget allocation (Question C1) to a 100% basis. In a small number of cases, particularly for the ranking exercise (Question B1), response patterns suggest that respondents might have reversed the order of entries (e.g. 1 as least rather than most important), but given the wide diversity of reported responses and plausible rankings, these were not altered.
Demographic questions used in this analysis are included verbatim in table 3, again with question ID included to facilitate interpretability of the supplementary data file.
As respondents often wrote custom comments about their jobs, political affiliation, and race in survey booklet margins rather than selecting a standard option, demographic data were extensively cleaned to match handwritten custom responses to presumed analytical categories as needed, a potential source of error. For example, respondents who entered 'German' as their race/ethnicity were coded as white, but US respondents who entered 'American' or Australian respondents who entered 'Australian' were coded as 'not specified.' Demographic summary data are published in (Grubert 2019), with population demographics reported in table 2 and sample demographics by mailing type reported in table 5 of that piece. Confirmatory analyses and hypotheses were preregistered through the Open Science Foundation and are presented in table 4 (they can also be viewed in detail online) (Grubert 2016). Most analyses and hypotheses concern regional and demographic variability in priority order associated with ranking and rating questions (B1, B2 and B3), which were primarily evaluated using a two-sided Kendall's tau correlation test at a significance level of α = 0.05. Although positive correlation is anticipated based on pilot data (Grubert 2017c), a two-sided test is a more conservative measure in the sense that statistical significance requirements are more stringent.

Results
Across the 34 confirmatory hypotheses preregistered for this analysis, 21 were supported by analysis (H1-H5; H9; H15-H20; H22-H25; H27-H30; H33). Values and statistics can be found in the supplementary data File. This main text emphasizes correlative results due to specific interest in evaluating whether respondents generally agree on higher and lower socioenvironmental priorities.
Supported hypotheses include the hypothesis that mean and median ratings for all 30 socioenvironmental considerations evaluated in questions B2 and B3 would exceed 3, 'moderately important' (H3; figure 1). Overall, the highest average ratings were for drinking water quality (4.7/5) and water supply (4.6/5). The highest average rating for an issue designated in the survey text as a social consideration (and the next highest average rating overall) was for healthcare (4.4/5).
The hypotheses that priority order for categories evaluated in questions B1-B3 would not vary by region (H6-H8) were incompletely supported (tables 5-7).
Priority order was statistically significantly correlated at the α = 0.05 level for nearly, but not all, regional pairwise comparisons. Exceptions for Question B1 (ranking) are: Pennsylvania and the Powder River Basin (τ = 0.38), Desert Sunlight (τ = 0.42), and Ivanpah (τ = 0.47); and the Powder River Basin and Genesis (τ = 0.42). Exceptions for Question B2 (environmental) are: the Powder River Basin and Australia (τ = 0.42). All region-to-region correlations were statistically significant at α = 0.05 for Question B3 (social). Similarly, hypotheses about variability in priority order for categories evaluated in questions B1-B3 based on respondents' personal characteristics (H22-H32) were incompletely supported (hypotheses were that there would be no difference between priority orders based on personal characteristics, except for political affiliation and self-identification as environmentalists). Priority order was statistically significantly correlated at the α = 0.05 level (and usually at the α = 0.001 level) for rankings based on health status (H22), happiness status (H23), gender (H25), education level (H27), employment status (H28), employment in resource-based sectors (H29), and financial security (H30), as hypothesized (i.e. there were no statistically significant differences among those reporting different values for those characteristics). Priority order was statistically significantly correlated at the α = 0.05 level for rankings based on political affiliation (H31) and self-identification as environmentalists (H32), contradicting the hypothesis that priority order would be significantly different for those characteristics. Hypothesis H26, that priority order would not vary based on race/ethnic identity, is incompletely supported: several pairwise comparisons were not statistically significantly correlated at the α = 0.05 level, but since 896/1311 respondents identified as white (with an additional 144 entering nonspecific custom data like 'American,' many of whom are likely white), the

ID
Scope Hypothesis

H1
Question B1, Entire sample Mean and median rank for human health will be less than or equal to 3 (i.e. human health will be a top-three priority on average) H2 Question B1, Entire sample Mean and median rank for culture will be greater than or equal to 8 (i.e. culture will be a bottom-three priority on average) H3 Question B2 and B3, Entire sample The mean and median of respondents' answers to B2 and B3 will be greater than or equal to 3 ('moderately important') for all rows: that is, no impact category will be considered 'not at all important' or 'slightly important' on average H4 Question B2 and B3, Entire sample The mean and median importance ranks for 'Water supply' and 'Drinking water quality' will exceed the mean and median importance ranks for 'Climate change' H5 Question C1, Entire sample Respondents will devote a mean of less than 33% of surplus budget to environmental improvement H6 Questions B1-B3, By subregion Differences in ranking based on mean values for impact categories in B1, B2 and B3 will not be statistically significant based on respondent's subregion (classified based on response to A1, zip code of residence), using Kendall's tau H7 Question B1, By subregion The mean priority order (taken as the ranked set of mean rank for each impact category) will be the same for each subregion, using Kendall's tau H8 Questions B2 and B3, By subregion The mean priority order for environmental impact categories (taken as the ranked set of mean rank for each impact category in question B2) will be the same for each subregion, using Kendall's tau H9 Questions B2 and B3, By subregion The mean priority order for social impact categories (taken as the ranked set of mean rank for each impact category in question B2) will be the same for each subregion, using Kendall's tau H10 Question C1, By subregion Respondents from the more isolated Tonopah * and Bakken regions will devote, on average, more than 33% of the budget surplus to social services ( * Crescent Dunes) H11 Questions A2, B2, Entire sample Mean scores for environmental impact categories (question B2) will be higher for respondents who respond 'yes' than for those who respond 'no' to question A2 about whether they consider the place where they currently live to be their home H12 Questions A2, B3, Entire sample Mean scores for social impact categories (question B3) will be higher for respondents who respond 'yes' than for those who respond 'no' to question A2 about whether they consider the place where they currently live to be their home H13 Questions A4, B2, Entire sample Mean scores for environmental impact categories (question B2) will be higher for respondents who report longer times lived in an area in question A4 H14 Questions A4, B3, Entire sample Mean scores for social impact categories (question B3) will be higher for respondents who report longer times lived in an area in question A4 H15 Question D8, Entire sample Retirees (D8) will be overrepresented as respondents relative to population demographics taken from census data for the appropriate zip codes H16 Question D5, Entire sample Women (D5) will be overrepresented as respondents relative to population demographics taken from census data for the appropriate zip codes H17 Questions B3, D1, Entire sample Mean values for importance of education in B3 will be higher for respondents who respond 'yes' to D1 about having children H18 Questions B3, D1, Entire sample Mean values for importance of community future in B3 will be higher for respondents who respond 'yes' to D1 about having children (Continued.) Questions B1-3, D9, Entire sample Differences in ranking based on mean values for impact categories in B1, B2 and B3 will not be statistically significant for those reporting a resource-based employment sector (manufacturing, mining and related services, oil and gas and related services, or utilities) versus other employment sectors in D9 (employment sector), using Kendall's tau H30 Questions B1-3, D10, Entire sample Differences in ranking based on mean values for impact categories in B1, B2 and B3 will not be statistically significant for those reporting 'not at all' or 'slightly' versus 'moderately,' 'very,' or 'extremely' secure in D10 (financial security), using Kendall's tau H31 Questions B1-3, D11, Entire sample Differences in ranking based on mean values for impact categories in B1, B2 and B3 will be statistically significant for democrats, republicans, and others based on responses to D11 (political party), using Kendall's tau H32 Questions B1-3, D12, Entire sample Differences in ranking based on mean values for impact categories in B1, B2 and B3 will be statistically significant based on responses to D12 (environmentalist identity), using Kendall's tau (Continued.) Question D13, Entire sample More than 2/7 th s of respondents will report finding the survey very or extremely interesting in D13, and fewer than 2/7 th s of respondents will report finding the survey very or extremely uninteresting in D13 (note: typo of 'D14' in place of 'D13' in preregistration) H34 Questions B2, B3, D13, D14, Entire sample People who leave substantive feedback (more than 'thanks!' or 'good luck!') in D14 will have more extreme opinions in B2, B3, and D13 than those who do not  Responses suggest that human health and the natural environment are particularly highly valued (figure 2). Sample-wide prioritization of drinking water quality and water supply persists at the regional level: drinking water quality and water supply were the top ranked environmental considerations for every region evaluated. Similarly, healthcare was the highest or second highest social consideration for every region evaluated except for the Powder River Basin. This survey's results confirm that climate change was an extremely polarizing issue in the studied areas at the time of study: across all 30 social and environmental categories, climate change did not rank in the top half in any region based on average rating (26/30 for the Table 6. Regional correlation in priority order for question B2, reported as Kendall's tau (τ is significant at 0.33 for α = 0.05, marked with * , and 0.58 for α = 0.001, marked with * * ; n = 15). Yet, the most commonly assigned rating (mode) for climate change was 5/5, or 'extremely important'-the only one of 11 impact categories with 5/5 as the modal rating, but with an average rating of less than 3.5/5 (3.3/5).

Limitations
Limitations of the survey instrument and fielding approach are extensively described elsewhere (Grubert and Skinner 2017, Grubert 2017a, noting particularly that intentionally low salience of the survey may have contributed to low response rates. Demographics in the surveyed communities were diverse (30%-100% white; 0%-51% below the poverty level; and across communities with widely differing levels of rurality) but not representative of national demographics. Neither the locations nor the responses are representative random samples, although this characteristic is not considered a major limitation for this exploration of empirical preference data in energy host communities. As with survey research in general, one major limitation of this study is that results are bound to the specific temporal context of response. The 2016 US presidential election in particular was noted in several free response comments. Quantitative ranking and rating responses are ordinal rather than cardinal data. Given the use of a paper survey, limited definitions were provided for response items, and item order was not randomized for the paper surveys. Survey data are fundamentally shallow data with limited contextualization that can pose challenges for theoretically rigorous interpretation, although this limitation was mitigated in part by use of short answer questions in the survey itself and deeper interview and observational data associated with the broader project (Drummond and Grubert 2017, Grubert and Skinner 2017, Grubert 2018b. Additional qualitative analysis of the short open-ended texts provided in this survey, and correlation with separate interview data, is a topic of future work that is expected to include both manual coding and natural language processing techniques to test consistency between quantitative survey responses and qualitative statements about place and priorities. Simultaneously evaluating primarily environmental and primarily social impact categories (figure 1) in sustainability assessment is a long-term, though challenging, goal for decision support tools (Guinée et al 2011). Rigorously and effectively measuring and evaluating impacts associated with outcomes that are difficult to quantify and/or have highly differentiated impacts based on timing, location, and affected parties is extremely challenging. As this exploratory work shows, though, respondents to this survey rate categories across the spectrum as important, with mean, median, and mode rating across the responses exceeding 3 'moderately important' for all 30 evaluated categories (H3). Extreme interpretive caution from this small and nonrepresentative sample is advised, particularly given limited statistical power for comparisons across racial and ethnic groups that experience highly disparate levels of impact from similar processes (Bullard 1999). Tentatively, however, this observation suggests that respondents largely agree that the socioenvironmental criteria that experts have identified as relevant for evaluating sustainability in methods like LCA are indeed important. It also reinforces the argument that primarily social impact categories, in addition to commonly used primarily environmental impact categories, are relevant for sustainability-driven decision processes (Jørgensen et al 2008, Benoît-Norris et al 2011. Considering all relevant impacts is a core tenet of LCA (ISO 2006). Particularly because understanding of what environmental considerations actually are varies (Van Liere and Dunlap 1981), with people from groups with lower levels of social power (e.g. lower wealth and racially minoritized people) tending to identify more socially oriented issues as environmental issues (Song et al 2020), a holistic approach to evaluating sustainability that explicitly incorporates value systems held by the most affected and most vulnerable people is needed to promote a just transition that emphasizes distributional justice and material well-being (Healy andBarry 2017, Grubert 2020).
Overall, results from this survey suggest that priorities for socioenvironmental outcomes associated with impact categories commonly used in environmental LCA and proposed for use in social LCA are relatively consistent across region and personal characteristics for this exploratory sample. This survey was limited to specific communities not designed to be representative of the broader population, so results should be interpreted as observations of diverse perspectives from specific geographically framed contexts rather than as statistically generalizable priority patterns (Raman et al 2015). Prioritization patterns also vary within regions, and clusters can be identified and analyzed using techniques like k-means clustering (Grubert 2017c). Although this study focuses on reporting the results of preregistered analysis related to inter-rather than intra-region patterns, individual project assessment within a given region could likely productively investigate value clusters to support discussion, visioning, and decision making. For example, communities might include groups with strongly held and highly incompatible visions of the future that might or might not be a result of highly differentiated priorities (Grubert and Skinner 2017). In high conflict situations where preferences across groups are similar, next steps for resolution and codesign might be quite different from situations where preferences across groups are very different. Nonetheless, the observation that there is general agreement on the most and least important categories is consistent with prior work focused on the general US population (Grubert 2017c). This finding, even without generalizability, suggests that the practice of equal weighting across socioenvironmental impact categories in tools like LCA does not achieve the frequently intended goal of removing value judgments from analysis and presenting neutral results. That is, if there is evidence that people consistently agree that some outcomes are higher priority than others, equal weighting will tend to overemphasize lower priority issues.
Within this context, one concern is that people might not express priorities that are consistent with scientific understanding of risk, threat, and impact severity. For this survey, it is notable that climate change in particular was not highly prioritized despite the extreme threat that climate change poses. This observation in itself raises important additional questions, like whether issue politicization interferes with people's feelings about actual outcomes, whether sustainability assessment is sufficiently granular (e.g. climate change is likely to exacerbate essentially all of the other environmental impacts considered within typical LCAs, which might not be well captured by a single-issue GHG metric: this is, to some extent, an issue with midpoint indicators that evaluate a driver of impact versus endpoint indicators that evaluate the impact itself (Huijbregts et al 2017)), and how globally versus locally affected people's priorities might differ. It is important to reiterate again that LCA is a decision support tool that informs human decision makers and thus has value only insofar as it is helpful, not a decision making tool that deterministically defines what should happen without additional intervention by people (Hunkeler 2006). The need to make potential value conflicts clear and explicit is not the same thing as using priority measurements deterministically. In general, though, some of the challenging issues raised by the observation that preferences might not correlate with understanding of severity are somewhat different methodological issues that LCA also needs to address. In particular, successful prioritization across impact categories fundamentally depends on normalization across impact categories, another serious and significant challenge for LCA. That is, LCA must be able to characterize how relatively severe an impact is in one category relative to another before preferences are invoked. Someone might prefer an apple to an equally good orange, but still select an excellent orange over a rotten apple. This issue is particularly salient for climate change, where impacts are potentially extremely severe and extremely widespread relative to potential impacts in other categories.
These major other challenges aside, given LCA's increasing status as an authoritative method in the context of energy policy design and other broad decision contexts (Safford 2022), the observation that people may agree on which issues are higher or lower priority is especially important as the community shifts from primarily performing snapshot attributional analyses for products and toward major prospective analyses of dynamic systems as inputs to policy. Particularly when future conditions are likely to be very different from past or current conditions, as with energy systems (Pehl et al 2017, Grubert 2021, LCA practitioners have a responsibility to identify and transparently express major sensitivities that could change decisions. Values, and particularly cross-category priorities, are among these major sensitivities. As such, extending work like that presented here to encompass more, and more diverse value systems, and focusing methodological attention on how to present these data in rigorous, replicable, and transparent ways that actually support decisions, is a major priority for LCA. The empirical preference data presented in this work do not substitute for best-practice participatory engagement, but rather serve as contextual preparation to support such processes (Bessette and Mills 2021). For example, observed patterns of expressed preferences, and observations about when such patterns are robust across place, demographic, and context, can be an input for decision support tools to show how different prioritizations might change decisions and what that means in a given context, or a reference for planners and policymakers to ensure they are prepared to engage on issues likely to be important in community consultation processes. Making preference data explicit can be a starting point for deeper conversations about values, what people wish to protect, and how projects or broader processes (e.g. climate change) might affect valued environments and relationships in relevant ways (Elser et al 2020).
Although the survey results presented here suggest generally high correlations in ranked priorities across regions and personal characteristics, which implies that priorities for topics commonly considered in sustainability assessment are not random, points of difference are instructive. The largest drivers of difference across regions explored in this work are related to regionally relevant needs and priorities, noting again limitations like the fact that both the respondents and the populations of the sampled places were often overwhelmingly white (tables 2 and 5, (Grubert 2019)). For example, considerations like supply of minerals like oil and coal, wages, and long-term employment were much more highly rated on average by respondents from the coal-dominated communities of the Powder River Basin than others, which likely reflects the fact that the survey coincidentally arrived in the region in the immediate aftermath of major layoffs from the local coal mines (Smith 2019). For the direct ranking question (B1), respondents from the solar communities rated governance as higher priority than other respondents, which is notable given the rapid expansion of energy infrastructure in the communities investigated. Polarizing issues like climate change and emergency social services were rated noticeably higher on average in the most urbanized region (Pennsylvania, with specific distribution in Pittsburgh and Philadelphia) and in Australia, where politicization patterns differ from those in the US (figure 2). Future work might focus not only on extending geographies of this approach, but also on investigating within-geography differences in priority patterns to identify points of local value conflict.
Although most rankings based on demographic group membership are statistically significantly correlated, as with the regional rankings, some of the specific discrepancies are informative. For example, the rank (based on aggregated ratings) of 'emergency social services like shelter and welfare' by respondents who self-identified as 'not at all' or 'slightly' financially secure, in good health, and happy was five or more positions ahead of its ranking by respondents who self-identified as 'moderately,' 'very,' or 'extremely' financially secure, in good health, and happy-an unusually large distance. Other large distances (more than five places different) include disparities in ranking for the importance of climate change (younger people ranked it as more important than older people; people identifying as Democrats or coded as left-leaning ranked it as more important than people identifying as Republicans or right-leaning), long term employment (younger people ranked it as more important than older people), and the supply of minerals like oil and coal (those who did not self-identify as environmentalists ranked it as more important than those who did).
In general, preference data are likely to be most valuable in sustainability assessment when they are used to support decision makers by making points of agreement and conflict visible across multiple preference scenarios, rather than as single, opaque weighting factors designed for deterministic, single-score evaluations. Points of contention across scenarios can identify needs for more in-depth, contextualized research (for example, see previously published, complementary qualitative work in some of the regions evaluated with this survey: (Drummond and Grubert 2017, Grubert and Skinner 2017, Grubert 2018b). Using preference data for scenario analysis intended to evaluate decision robustness to different value systems (Grubert 2017c), or to identify issues that have become markers of group identity or otherwise taken on meaning beyond their specific role (Fischhoff 2013), can inform quantitative sustainability assessment without being fully prescriptive. Particularly because evidence suggests that people often roughly agree on higher versus lower priorities, using empirical data elicited from real people can help identify likely points of real contention. Understanding the energy system as linked physical and societal infrastructure, and acknowledging local priorities, can contribute to better decision making and a more sustainable future.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary information files).