Drilling Setbacks vs Government Takings: The Case of Colorado’s 2018 Colorado Ballot Initiatives

Abstract

State policymaking is at the center of many oil and gas related disputes. Driven by the promise of affordable energy, economic development, and new revenues, supporters of oil and gas have pushed for policies designed to nurture the oil and gas industry, whereas opponents have emphasized a myriad of environmental threats and disruptions. Statewide ballot initiatives related to oil gas extraction offer a particularly useful lens to examine the preferences of voters in that states are home to residents who directly and indirectly experience the promises and perils of extraction. This paper examines two ballot initiatives in Colorado from 2018, one of which was supported by the oil and gas industry; the other of which was opposed by the same industry. We find that the inclusion of natural amenities, livelihood, and population change provides a useful set of variables for further study.

Keywords:

• Land use
• livelihood
• natural amenities
• oil and gas policy
• resource extraction
• takings

Notes

1 The question of “regulatory taking” arises when legislation or regulation, or the implementation of that regulation, denies the property or landowner of all “economically viable use of his land” (Renaud 2020).

2 Setbacks are an umbrella term that can be described as the required distance between oil and natural gas facilities (or other buildings/facilities) and various buildings/resources such as homes, schools, bodies of water, or areas of outdoor recreation (Fisk 2017).

3 The literature on ballot measures finds that the more complex the language of the initiative, the higher the level of ballot roll-off (Reilly and Richey 2011). Proposition 112 (setbacks) contained the longest and most complex language compared to Amendment 74 (takings). However, the roll-off was greater for Amendment 74 (2,454,387 total votes) than Proposition 112 (2,488,022 votes). We suspect that the very large media campaign funded by the oil and gas industry in Colorado ($32 million spent, of which $31 million was spent on the ‘no’ campaign against 112; Ballotpedia 2018a, 2018b), overwhelmed any roll-off that would normally be expected because of ballot language.

4 Previous research has also noted a connection between gender and oil and gas attitudes (Malin, Ryder, and Hall 2018; Howell et al. 2019, O’Hara et al. 2015). However, there is not sufficient variation relative to gender between counties to include in our analysis.

5 We also summed oil spill data (Colorado Oil and Gas Conservation Commission (COGCC) 2020) by county between 2014 and 2018, but this variable proved to be a poor predictor of voting on Prop 112 and Amendment 74.

6 Another possible indicator of natural amenities is the percentage of employment in hotel and restaurant industry, which in Colorado is likely to be an indicator of resort activity. This indicator was not a strong of a predictor of voting on Prop 112 and Amendment 74.

7 For all MGWR analyses, we used MGWR Version: 2.2.1, Released on: 03/20/2020, source code is available at: https://github.com/pysal/mgwr, Development Team: Ziqi Li, Taylor Oshan, Stewart Fotheringham, Wei Kang, Levi Wolf, Hanchen Yu, Mehak Sachdeva, and Sarah Bardin, Spatial Analysis Research Center (SPARC), Arizona State University, Tempe, USA.

8 The MGWR approach takes advantage of spatial scale and context to derive estimates of the number of nearest neighbors (bandwidths) over which social “processes exhibit spatial heterogeneity” (Fotheringham, Li, and Wolf 2021, 1607). Based on the bandwidths, MGWR estimates standardized slopes and confidence intervals for each county. The bandwidths are a crucial component for estimating how spatial scale and context influence votes on these two initiatives among Colorado counties. For additional details about MGWR and superior capacity to predict outcomes compared to GWR and OLS see Fotheringham, Yang, and Kang 2017; Fotheringham, Li, and Wolf 2021.

9 Political Geographers often focus on spatial variables to explain voting patterns among counties (Li et al. 2020). We do not take the perspective of Gary King (1996) who argues that spatial considerations should be moot if each geographic unit is properly identified by the geographic variables in the analysis.

10 In MGWR, each Beta for each covariate in each county has its own level of significance. The adjusted T-values that are reported in Tables 2–4 are calculated to account for multiple hypothesis testing and dependency which is the reason they are different from the usual t value around 1.96.

11 This outcome could also be a function of using the county as the unit of analysis. However, an analysis of voting precincts found a negative relationship between proximity to oil and gas operations and voting for 112 (Raimi, Krupnick, and Bazilian 2020).

12 We also conducted this OLS with 1000 bootstrap samples. The bootstrapped bias estimates showed no substantive bias that would change our interpretation of the OLS estimates. We do not report the bootstrap bias estimates in Tables 3 and 4.