**1. Introduction**

Natural gas pipelines extend through every state in North America to connect producers, distributors, and customers. Proposals to construct, expand, or repurpose pipelines often lead to contention over risks to host communities. Examples include recent debates over the Mountain Valley, Atlantic Coast, and Tennessee Gas pipelines. Some communities have enacted policies to deter natural gas pipelines [1], while others have welcomed them [2]. Decisions about pipeline construction and regulation are often made with scant information about the risks and costs for host communities. If well informed, prospective host communities can weigh the risks associated with natural gas transmission against the long-term benefits. This article provides information to assist communities and pipeline operators with the appropriate cost–benefit analysis, and o ffers possible remedies for the problems communities face regarding risk spreading and uncertainty.

In 2019, pipelines supported annual expenditures of almost \$150 billion on natural gas in the United States for the production of heat, electricity, plastics, fertilizers, pharmaceuticals, fabrics, and organic chemicals, among other uses [3]. Figure 1 shows the steadily increasing use of natural gas in the United States. Figure 2 shows the percentage of natural gas used for each purpose. With the benefits of natural gas transmission come the threats of damage to life and property. After the construction phase of pipelines, the external costs stem largely from leaks or the combustion of toxic loads and the resulting damage to property, health, and the environment.

**Figure 1.** U.S. natural gas consumption, 2000–2019. Data source: U.S. Energy Information Administration.

**Figure 2.** U.S. natural gas consumption by sector, 2019. Data source: U.S. Energy Information Administration.

The Pipeline and Hazardous Materials Safety Administration (PHMSA) reports a total of 12,316 natural gas, hazardous liquids, and liquefied natural gas pipeline incidents between 2000 and 2019 [4]. The repercussions included 309 deaths, 1232 injuries, and \$10.96 billion in property damage. These figures are accessible, ye<sup>t</sup> a majority of the underlying incidents are irrelevant to communities that might host a natural gas transmission pipeline. Some of the incidents occurred offshore, some involve more volatile substances than natural gas, and some occurred in gas gathering and distribution operations that stem from a different set of decisions than transmission pipelines.

This article focuses on the information relevant to prospective host communities for natural gas pipelines. Using data paired down to include only onshore natural gas transmission pipeline incidents, the article provides incident rates and estimated costs of bodily injury, lost life, and property damage. Regression analysis provides further insights into expected damage costs based on community and pipeline characteristics. The article also discusses approaches to risk managemen<sup>t</sup> for communities that could be applied in any country.

Section 2 of this article provides a review of the related literature. Section 3 explains the methods used to establish the dataset and estimate the regression coe fficients. Section 4 reports the results. Section 5 discusses implications and possible remedies for host communities' exposure to risk and uncertainty. Section 6 concludes the paper.

#### **2. Literature Review**

The previous literature on the costs of pipelines to host communities focuses largely on the e ffects of pipelines on property values [5]. Reductions in property values near pipelines reveal perceptions of the risks of leaks and explosions. If consumers were informed, rational, and risk-neutral, the loss in property values would accurately reflect the expected cost of such incidents, and it would be redundant to add pipeline-related decreases in property values to the costs of property damage, injuries, and deaths when calculating the total cost of pipeline incidents. If consumers have imperfect information, the e ffect of pipelines on property values is not an accurate measure of the expected cost. Residents who perceive no risk of pipeline incidents are willing to pay the same amount for a home regardless to its proximity to a pipeline.

The findings on pipelines' e ffects on property values are inconclusive. Studies by McElveen et al. [6] and Integra Realty Resources [7] are among those suggesting that pipelines have no significant influence on property values. In contrast, Simons et al. [8] and Hanson et al. [9] estimate that major incidents involving oil and gas pipelines lower property values by 10.9%–12.6% and 4.65%, respectively. Kielisch [10] provides evidence from realtors, homeowners, real estate appraisers, and land sale analysis that natural gas pipelines can lower property values significantly, and in some cases by as much as 39%. Herrnstadt and Sweeney [11] point out that accurate information on pipeline risks would allow people to respond with appropriate safety plans. Another benefit of the present research is that it provides information with which homebuyers can make better decisions about their willingness to pay for homes near pipelines.

Another body of research presents models of risk assessment for pipelines. That research allows operators to fine-tune their risk estimates based on situation-specific characteristics such as the density and pressure of gas within the pipeline [12–14]. The present research incorporates broader community characteristics such as mean income and population density, along with the age of the pipeline, as determinants of the cost of an accident. The latter determinants are relatively constant and readily available to communities considering the prospect of a new pipeline.

Economists must reluctantly place a value on human life to inform decisions about tradeo ffs between money and lives, including decisions about safety regulations, environmental policies, and pipelines. The existing literature addresses the value of unidentified or "statistical" lives such as the lives that could be lost by a community hosting a pipeline. We know that statistical lives have finite value because communities make decisions that have finite benefits and involve risk to life. By allowing people to drive cars, cross streets, operate farm machinery, smoke, and use natural gas, it is inevitable that deaths will result. If the value of a statistical life were infinite, none of these activities would be acceptable. Estimates of the value of a statistical life come from real-world tradeo ffs people make between money and risks of death as revealed in labor markets among other settings. In a recent synthesis of the available research, Viscusi [15] estimated that the bias-corrected mean value of a statistical life is \$10.45 million. Several U.S. agencies apply similar estimates, including the Occupational Safety and Health Administration, the Food and Drug Administration, and the Environmental Protection Agency. A related vein of literature exists for the value of bodily injury. Viscusi and Aldy [16] provide a summary of 24 relevant studies of the value of a statistical injury, the

mean of which is \$90,697. These values for a statistical life and a statistical injury are applied to deaths and injuries in the present study.

#### **3. Data and Methods**

Data on natural gas pipeline incidents are available from the Pipeline and Hazardous Materials Safety Administration (PHMSA), a division of the U.S. Department of Transportation [17]. The PHMSA dataset o ffers information on every reported natural gas pipeline incident in the United States, including the location of the incident, the cost of property damage, the number of injuries and deaths, and the age of the pipeline. The PHMSA requires that incidents be reported if they cause a death or in-patient hospitalization; at least \$50,000 in property damage excluding lost gas; the unintentional loss of at least 3 million cubic feet of gas; an emergency shutdown of an underground natural gas storage facility; or an event that is significant in the judgment of the operator, even if it does not meet the other criteria [18].

Natural gas pipeline incidents involve both explicit and implicit costs. The explicit costs include the costs of public and private property damage and emergency responses, all of which are reported to the PHMSA. The implicit costs are the costs of injuries and lost life, estimated by multiplying the number of injuries and deaths by the value of each type of occurrence drawn from the literature on the value of a statistical injury and life [15,16].

For the regression analysis, those pipeline data were paired at the zip-code level with information from the U.S. Bureau of the Census on population density, and information from the Statistics of Income Division of the U.S. Internal Revenue Service on income, real estate taxes, and the percentage of tax returns that are farm tax returns. The population data come from the 2010 census, conducted halfway through the 2000–2019 time period being studied. The tax data came from 2017, the most recent year for which complete data were available. Table 1 provides variable definitions for the dataset.


**Table 1.** Variable definitions and summary statistics.

The selected variables represent location characteristics that could influence the consequences of a pipeline incident. In related regression analysis of property damage from hazardous liquid pipeline incidents, Restrepo et al. [19] use a dummy variable for high-consequence areas, which include areas with high population density. The present research uses population density among other values that similarly a ffect incident cost. The specification was subject to the availability of data. It would be ideal to have measures of the population density and the value of real estate within close proximity of the pipeline. Actual data are available at the zip code level, which is not always limited to areas in close range of the pipeline. The specification was adjusted in response to empirical findings on the contribution of particular variables, as discussed further below.

The population density, mean income, and real estate taxes per square mile could each influence damage costs positively or negatively. A higher population density could increase the likelihood of a leak or explosion being near buildings and people. At the same time, areas with high population densities can have stricter requirements for pipe strength, stress levels, or monitoring, which decrease the likelihood of a high-cost incident [20]. Higher mean income similarly increases the likelihood that an incident of any particular scale would cause costly damage, but correlates with greater protections against major incidents. For example, Pless [21] reports that the U.S. state with the lowest mean income, Mississippi, had 50.7 inspection person days per 1000 miles (1609 km) of natural gas transmission pipeline in 2009, whereas the U.S. state with the highest mean income, Massachusetts, had 764.3. Controlling for population density and mean income, having higher real estate taxes per square mile is hypothesized to have a positive influence on the cost of property damage because, for any given tax rate, it rises with property values.

Incidents along gathering and distribution lines are not included in this research, because they result from a di fferent decision-making process than transmission lines. The risk of incidents along transmission lines is an inherent aspect of playing host for the natural gas industry as it brings its product to distant markets. In contrast, distribution lines are the result of consumers in each municipality deciding to use natural gas as fuel. Further, many of the incidents in the distribution pipeline category occur at customers' homes and businesses. Gathering lines are in a distinct category as well. They are part of the natural gas production process and serve the purpose of bringing fuel from the extraction site to a central collection site. O ffshore pipeline incidents are not included, because they are not related to the issue of communities hosting transmission pipelines.

The primary equation used to estimate the determinants of pipeline incident costs is

ln*Costi* = α0 + β1*Regioni* + β2*Population Densityi* + β3ln*Mean Incomei* + β4ln*Pipeline Agei* + β5ln*Real Estate Taxesi* + *i*

*Region* is a vector of the *East*, *Midwest*, and *South* dummy variables. The *West* dummy variable is omitted to avoid multicollinearity. Zip codes starting with 0–2 are in the East, those starting with 4–6 are in the Midwest, those starting with 8 or 9 are in the West, and those starting with 3 or 7 are in the South. The dummy variable *Midwest* is used instead of *North* because the observation level is zip-code areas, which are numbered from east to west. Zip-code areas starting with 8 and 9 run from the northern border to the southern border of the United States. It is, therefore, more practical to delineate the Midwest and West regions. The empirical investigation included several variations of this equation to assure the robustness of the findings.
