The effect of industry activities on public support for ‘fracking’

ABSTRACT

Research suggests that previous, current, and prospective extractive industry activities influence perceptions of new development. Studies that have drawn this conclusion, however, have usually focused on specific projects in specific communities. Here, these factors are examined on an aggregate, national scale. Combining geospatial data on extractive industry activities and survey data from a nationally representative sample (N = 1061), the influence of extractive industry activities on support for fracking is studied. While limited evidence is found for the impact of proximity to oil and gas wells or production on support for fracking, employment levels in the natural resources and mining sector in the respondent’s county and residence in an area experiencing active oil and gas development significantly increase support for fracking. The results highlight the role of spatial and community factors in shaping support for energy development.

KEYWORDS:

• Hydraulic fracturing
• natural gas
• energy
• shale
• unconventional oil and gas development
• GIS

Acknowledgments

The authors would like to thank the reviewers and editor for their helpful comments. The Climate Change in the American Mind survey was funded by the Surdna Foundation, the 11th Hour Project, the Grantham Foundation, and the V.K. Rasmussen Foundation.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. We acknowledge that the term ‘fracking’ has become both politically charged (Clarke et al. 2015) and used to represent a range of topics related to the process, including its impacts and the oil and gas industry generally (Evensen et al. 2014). However, we choose to use it here to maintain consistency with the terminology used in our survey.

2. Our measure of support/opposition toward fracking is fundamentally different from measures of risk perceptions or attitudes. Unfortunately, we did not have space in our survey to ask additional questions of this nature. One attitudinal question did ask whether the respondent thought fracking is a ‘good or a bad thing’ on a scale from +3 to −3. Responses to this question were highly correlated with fracking support/opposition, indicating that at least attitudes may be closely aligned with support/opposition.

3. When the survey was administered, the ‘undecided’ response was indicated to be a neutral position between ‘support’ and ‘oppose’. For this reason, we have included it here. Respondents saw these options as a continuum on the survey. In contrast, ‘refused’, ‘prefer not to answer’, and ‘do not know’ options were outside of this continuum and thus dropped from our analysis.

4. Research on siting disputes has shown that the distance within which a siting controversy generates noticeable perception-level impacts varies. Lober (1995) found a perception-impact area of no more than one mile from the disputed site. More recently, the concept of a circular buffer of 20 kilometers (12.4 miles) has been used in energy siting studies (Braunholtz 2003, Warren et al. 2005, Swofford and Slattery 2010). Stoffle et al. (1991), in their assessment of a ‘risk perception shadow’ of a nuclear waste facility, found that perceptions were most strongly affected within a 15-mile ‘core’ but influence could be detected as far away as 35 miles. Considering the relatively consistent findings from several research efforts on the effect of proximity on public perceptions toward energy facilities, we chose a distance of 15 miles as the appropriate distance from within which to measure historical oil and gas industry activity.

5. To maintain confidentiality, the Census Bureau provides ranges of total employment if only one business exists in a county in a sector. We replaced this range with its mean for all categories except the highest ‘100,000 or more’ employees, which was kept at 100,000. We chose to use natural resources and mining sector employment, as opposed to oil and gas sector employment specifically, because it limited the number of mean replacements due to confidential data.

6. In addition to ordinal logistic models presented here, we conducted a number of robustness checks using other model specifications. First, to allow for variations in other place-based, county-level characteristics not specified in our model, we ran a linear mixed-effects model with county as the upper level and individuals as the lower level, as suggested by scholars who combine individual and place-based characteristics (Hamilton et al. 2010). Compared to the ordinal regression models, the mixed effects models achieved identical signs and similar significance for all covariates. To account for the ordered categorical nature of the dependent variable, we chose to report the ordinal logistic regression model results. Next, as a robustness check for the ordinal regression models, missing value imputation and alternative categorizations of the dependent variable were conducted. Neither produced reversals in sign or substantive changes in significance compared to the results as presented.