**4. Results**

Looking only at the data on onshore natural gas transmission pipelines, between 2000 and 2019, there were 1846 incidents, 49 deaths, 173 injuries, and \$1.7 billion in property damage. Of the 12,316 total pipeline incidents reported in the introduction, only 15% were along onshore transmission pipelines, which shows the importance of breaking out this category of incidents. Table 2 separates these figures by region for the most recent decade. The West had the most deaths, the most injuries, and the most property damage, despite having the second-lowest number of incidents. The coe fficients on the regional dummy variables in the regression findings below support this finding and provide further insights into regional di fferences.


**Table 2.** Incidents and damage by region, 2010–2019.

There are about 300,000 miles (482,803 km) of onshore natural gas transmission pipelines in the United States, and there were 115 incidents in 2019. Table 3 shows the number of incidents per 10,000 miles (16,093 km) of these pipelines over the past 20 years. The numbers are notably consistent, with a mean of 3.11 and a standard deviation of 0.596. This indicates the relative predictability of incidents on a national scale and the inability of current safety regulations to eliminate risks.


**Table 3.** Incidents per 10,000 miles (16,093 km) of pipeline, 2000–2019.

a The PHMSA figure for miles of onshore transmission pipelines for natural gas in 2019 is not ye<sup>t</sup> available, so this figure was estimated using the miles of pipelines in 2018.

Table 4 provides the results of the primary regression. Except for the *Midwest* and *South* dummy variables, the effects of these variables are statistically significant at the 95% confidence level. Population density and the natural logs of mean income have negative and significant coefficients, showing that the influence of these variables on safety precautions dominates the influence of population density and income on the proximity of people and buildings to the pipeline, as discussed in Section III. This is the case holding constant the real estate taxes per square mile, a gauge for the value of property in the area. When the real estate tax variable is removed, as shown in Table 5, the significance of mean income falls, perhaps because mean income becomes a proxy for both more inspections (a negative influence) and more valuable property (a positive influence).




**Table 5.** Additional regressions (dependent variable: Cost of incident. t-values are in parentheses).

The negative coefficients on *East*, *Midwest*, and *South* were expected given the relatively large cost of incidents in the West, as apparent from Table 2. As hypothesized, the coefficient on the *Log real estate taxes* variable was positive and significant at the 95% level, showing that in areas with relatively valuable real estate, and thus larger yields for real estate taxes, an incident causes more costly damage.

The coefficient on *Log pipeline age* indicates that a one percent increase in the age of a pipeline corresponds to a 0.222% increase in the expected cost of a pipeline incident. Applying that to the average cost of a pipeline incident, a one percent increase in pipeline age represents an increase of \$3293 in the cost of the average incident along that pipeline.

The influence of pipeline age on damage costs is relevant to potential host communities for several reasons. To the extent that newer pipelines are safer than older pipelines, new projects have lower expected damage costs than existing projects. The rate of decline in pipeline safety over time is also relevant to communities as they consider the prospect of incidents well into the future, when the character of the community and its level of development may change. In addition, the risks associated with old pipelines must be considered when new projects involve the repurposing of existing pipelines. The average year of installation for a pipeline involved in an incident since 2000 is 1973.

Overall, the regression analysis reveals the indiscriminate nature of damages from pipeline incidents. The *<sup>R</sup>*-squared indicates that 7.4% of the variation in costs is caused by the variables in the equation. So even factoring in the influence of these variables, there is considerable uncertainty about the cost imposed by a leak or explosion. The largest sources of variation are specific to individual cases and are not captured by the variables in the dataset. This motivates communities' need for additional forms of insurance to mitigate risk and uncertainty, as discussed in Section 4.

Several versions of the regression equation were estimated to test the robustness of the findings. Table 5 shows the results. The signs on the coefficients and their significance are largely consistent with a few exceptions. Regression 1 repeats the findings discussed above for the purpose of comparison. Regression 2 is a linear version of the specification, which is an inferior fit but demonstrates the robustness of the findings. Regression 3 substitutes the percent of farms for the real estate taxes per square mile. Like the real estate taxes variable, the percent farms variable has a negative coefficient and is significant at the 95% level. Using both of those variables lowers the adjusted *<sup>R</sup>*-squared and causes both variables to lose their significance at the 95% level. Regression 4 includes neither real estate taxes nor percent farms, yielding results similar to the other regressions but an inferior fit. Regression 5

provides coe fficients for the specification with a dependent variable of property cost only, rather than total cost.

## **5. Discussion**

The results indicate that incidents along onshore natural gas transmission lines represent a small fraction—15%—of all pipeline incidents reported to the PHMSA between 2000 and 2019. Over the past decade, an average of 101 such incidents occurred annually in the United States. When an incident occurs, the damage can devastate the local community. Compensation for lost lives, bodily injuries, property damage, environmental damage, and related expenses are often subject to litigation. Pipelines can also create fears and anxieties in communities that go uncompensated. The findings of this research give communities a better idea of the scale and frequency of relevant incidents and quantify identifiable influences on damage costs.

Current approaches to pipeline safety focus on regulation. For example, in response to deadly incidents along onshore gas pipelines, the PHMSA tightened its integrity managemen<sup>t</sup> requirements in 2019. The new rules require pipeline operators to take further precautions, such as additional monitoring of the pressure in natural gas transmission pipelines and more assessments of pipelines in areas that are populated but not designated as high-consequence areas [22]. Such regulations are valuable attempts to increase pipeline safety, but they do not assist the victims of pipeline accidents when, despite the regulations, they occur.

To serve both pipeline companies and host communities well, policymakers must attend to the dual realities of low-incident probabilities and high costs for the rare victims. Solutions should also address the troubling uncertainty for pipeline hosts. The five worst onshore natural gas transmission pipeline incidents over the past decade each caused more than \$25 million worth of property damage [4]. All damage is disruptive to the property owners and victims, whether compensation is provided or not. Incidents involving explosions generate inordinate media attention and corresponding fears and concerns.

Communities would benefit from the certainty of insurance against the downside risk of a pipeline leak or explosion. One solution is for pipeline operators to act as insurers. These firms have a relatively clear understanding of the risks. They also make decisions that influence the pipelines' safety, meaning there are beneficial incentive e ffects of pipeline operators serving as insurers. Operators could budget for the expected cost with knowledge of the pipeline's history, the safety measures in place, and the monitoring practices, and they provide certain compensation when problems occur.

The pipeline operators are able to spread the risk of a costly incident across their entire pipeline network. In his concurring opinion on the legal case of *Escola* vs. *Coca Cola* [23], Justice Roger J. Traynor remarked on the ability of companies such as Coca Cola to spread the risk of injuries caused by their products broadly as a cost of doing business. Already, many of the costs of pipeline incidents are covered by the pipeline operators as payments to communities and damage awards in litigation. If full compensation of host communities became mandated or contractual, the pipeline owners would provide certainty where it is needed. In addition, by internalizing the external costs of their decisions, baring other sources of market failure, firms would make socially optimal decisions about pipeline construction and use.

Personal injuries have explicit and implicit values and require special consideration. In the case of lost human life, no amount of ex-post compensation is enough. However, we can apply the value of a statistical life—the life of an unidentified individual whose death we can anticipate due to the risks inherent in pipeline use. The literature review explains that the estimated value of a statistical injury is \$90,697 and the estimated value of a statistical life is \$10.45 million. If guaranteed in advance, compensation at these levels would provide appropriate incentives for firms and fitting ex-ante assurance for communities.

One form of the insurance remedy would be an application of the precautionary polluter pays principle [24]. The pipeline owners could create a trust fund with the amount that would compensate the victims for the worst-case scenario. That amount would go to the community in the event of damage and the trust fund would be replenished. If no damage occurred over the lifetime of the pipeline, the money in the fund would be returned to the pipeline owner. The results of this study regarding the expected cost and the influence of pipeline age, among other variables, would be informative for any such solution.
