Summary

  • A convergence of aging physical infrastructure operating past design life and weakening institutional oversight has narrowed the margin between normal operations and catastrophic failure at U.S. chemical facilities.
  • The average U.S. chemical plant is 46 years old — more than double the roughly 20-year useful life of its equipment — while insurance analyses document mounting losses as plants operate beyond the 30-year mark.
  • Regulatory coverage gaps leave major hazardous chemicals outside the EPA’s Risk Management Program, producing a two-tier safety architecture in which a significant share of recent deadly incidents occurred at facilities not subject to the agency’s most stringent rules.
  • A 25 percent single-site repeat-incident rate across the 600-plus events reported to the CSB since 2020 indicates that post-incident corrective actions are not preventing recurrence.
  • Bipartisan political failure has produced a policy environment in which rules finalized in 2024 to close oversight gaps are proposed for permanent shelving and the federal body that investigates root causes faces defunding.

Serious chemical accidents in the United States climbed to 131 in 2025, a 20 percent increase over the prior year, killing 48 people — nearly double the 2024 toll — and seriously injuring 142, according to a Wall Street Journal analysis of data submitted to the U.S. Chemical Safety and Hazard Investigation Board. More than 600 serious incidents have been reported across 46 states since mandatory reporting began in 2020. The sharpest escalation is structural: the physical infrastructure is aging past design life while the institutional mechanisms meant to detect and correct for that aging are being weakened and politicized. The convergence, not any single failure point, is what makes the current trajectory fragile.

The Physical Substrate Is Operating Past Design Life

Former CSB and Environmental Protection Agency officials, as well as industry safety experts, attribute much of the rise in accidents to years of deferred safety maintenance on aging plants, the Journal reported. Mike Schmidt, principal at Bluefield Process Safety, placed the average age of a U.S. chemical plant at 46 years; equipment typically has a useful life of roughly 20 years. Marsh, the insurance broker and risk advisory firm, estimated the average age of North American refining and petrochemical infrastructure at 61 years and has documented mounting losses as plants go beyond the 30-year mark. CSB reports have cited corroded pipes and absent safety equipment among the avoidable failures behind individual incidents.

The consequences of that aging were visible in two major incidents in May 2025. At a Nippon Dynawave Packaging Company plant in Longview, Washington, a 900,000-gallon tank containing white liquor — a corrosive chemical used in papermaking — collapsed, killing 11 workers. The facility had a documented history of state safety violations and worker complaints, including one filed weeks before the accident after an employee fell into a sinkhole of scalding pulp stock due to a maintenance oversight; the company characterized that event as a “near miss,” state labor records show. Days earlier, a chemical tank at a GKN Aerospace facility in Garden Grove, California — a 60-year-old plant — overheated, forcing the evacuation of more than 40,000 residents from a densely populated area near Los Angeles. GKN executive Steve Carlin attended a June 9 community meeting at which the company addressed the ongoing investigation into the incident.

The aging of chemical infrastructure exhibits what Nassim Nicholas Taleb has characterized as a concave response profile with respect to operational stress. As components age past design life, corrosion accumulates, and deferred maintenance compounds, the margin between normal operating load and catastrophic failure narrows. A pipe that has operated safely for decades does not degrade incrementally toward failure — it fails when accumulated fatigue reaches a threshold, and the consequence is disproportionate to the proximate trigger. When capital stock ages, the annual outlay required to maintain safe function rises, but the benefits of that spending — an accident that does not happen — are invisible on a balance sheet. The immediate, visible gain comes from delaying the expenditure, creating a concave payoff structure: each year of postponed maintenance yields a small, certain cost saving while the probability of a catastrophic failure, though individually low, increases monotonically, and the associated cost — in lives, litigation, and reputational damage — can be existential for the operator. Some of the increase in reported incidents may also reflect a rebound in industrial activity following the Covid-19 lockdowns, the Journal noted.

Roughly a quarter of the 600-plus incidents reported to the CSB since 2020 occurred at sites with at least one prior report — a repeat rate that signals the post-incident corrective-action process, whether company-internal or regulatory, is not preventing recurrence. Kinder Morgan, a $70 billion pipeline company, tallied 28 incidents across 26 different facilities over six years. The company stated it takes “every incident, large or small, seriously, and we are committed to continuous improvement and risk reduction.” Tyson Foods has reported eight incidents since 2020, including a fatal explosion caused by a ruptured oil hose at a Camilla, Georgia, facility in 2024, which resulted in an Occupational Safety and Health Administration citation for a serious violation. Neither Nippon nor Tyson returned requests for comment, the Journal reported; both have previously expressed that they prioritize worker safety.

The Regulatory Architecture Is Split and Weakening

The EPA subjects facilities that work with a list of regulated chemicals to stringent safety rules under its Risk Management Program. Currently about 11,500 facilities hold those chemicals above threshold quantities. But many toxic and deadly chemical combinations are not on the list. White liquor, the corrosive substance involved in the Washington disaster that killed 11, is not covered. Methyl methacrylate, the toxic chemical used in making plastic that triggered the Garden Grove evacuation of more than 40,000 people, is not covered either. Facilities handling such substances must observe certain basic safety standards but are not subject to RMP audits, safer-technology analyses, or worker stop-work authority. The EPA stated that the Clean Air Act requires any facility handling extremely hazardous substances to safely manage them, regardless of RMP status. Whether white liquor and methyl methacrylate were formally evaluated and excluded from the RMP list, or have simply never been evaluated, is unavailable from the material reviewed.

The result is a regulatory landscape with two tiers: a monitored tier where rules may be tightening depending on political cycles, and an unmonitored tier where aging infrastructure operates with fewer external checks. The Biden EPA in 2024 finalized rules designed to close some of these gaps — a post-incident third-party audit requirement that would create an external accountability loop absent at non-RMP facilities; a safer-technology analysis requirement; and worker-safety stop-work authority that shifts operational control to the workers closest to the hazard, a direct intervention against the deferred-maintenance dynamic. Those rules have not fully taken effect. The Trump EPA is proposing to permanently shelve them.

The EPA’s defense rests on two claims. First, that the CSB data the Journal analyzed includes facilities not governed by RMP rules, and that for RMP facilities, the agency’s own data show accidents declining between 2014 and 2023 — a comparison window that ends before the period of sharpest increase documented in the Journal’s 2025 data. “Accident prevention was and continues to be a priority for EPA in the Trump Administration,” the agency stated. Data scientists from environmental and worker-safety organizations counter that the Trump EPA’s analysis misses hundreds of incidents identified through news reports and other records, in part because facilities not under RMP jurisdiction face no timely reporting obligation. The Trump EPA’s argument that CSB data overstates the problem because of jurisdictional gaps inadvertently describes the problem it is cited to minimize: the gaps are real, and the dataset the EPA uses to demonstrate improvement excludes the very universe of unregulated facilities where some of the worst recent incidents have occurred, including the Longview and Garden Grove disasters.

The coverage gap produces systematic under-visibility. The Coalition to Prevent Chemical Disasters, which tracks news reports of chemical incidents, counts roughly twice as many incidents as the CSB between 2021 and 2025, because the CSB does not require reporting of accidents that resulted only in shelter-in-place or evacuation orders. The American Petroleum Institute’s industry survey data apply yet another set of member-reported criteria, generating a divergent picture of the same physical reality. Three overlapping but non-identical datasets produce three different accounts of whether the problem is growing or shrinking — a fragmentation that weakens the information base on which corrective policy depends.

The Political Feedback Loop Undermines Corrective Capacity

Current and former government officials close to chemical policymaking blame a bipartisan failure across Republican and Democratic administrations to adequately address the problem, the Journal reported. The CSB, which lacks enforcement power but performs root-cause investigations — a function advocates and industry safety officials alike have described as critical — has been targeted for defunding by the Trump administration since its first term, on the grounds that it duplicates OSHA and EPA work. Trump administration officials have argued that the CSB is unnecessary; advocates and industry officials alike counter that the agency, though understaffed and lacking enforcement authority, has been a crucial watchdog conducting autopsies of catastrophes.

CSB board members told the EPA the proposed rollback of the 2024 rules would be “a significant step backwards.” The EPA responded by saying, “CSB is clearly trying to prove they’re needed,” noting the two current board members are “holdovers from the Biden Administration.” The institutional friction introduces delay in any regulatory response, allowing the underlying physical deterioration to continue. Defunding the CSB would remove the one agency whose mandate is to investigate what went wrong without imposing penalties — an information-producing function no other agency conducts.

Consequences Fall Asymmetrically on Communities

Stakeholders bear sharply unequal consequences. Workers inside the plants face direct physical harm: the 48 fatalities and 142 serious injuries in 2025 occurred at the point of production. Residents of surrounding communities bear the exposure risk of chemical releases. Studies cited by the Journal indicate such incidents can depress home values for more than a decade and increase risks of cancer and respiratory illness. The EPA estimated in 2016 that approximately 177 million Americans lived within worst-case-scenario zones for RMP-regulated facilities alone — a figure that excludes the universe of non-RMP chemical operations. According to 2014 reports from watchdog groups that accessed catastrophe planning documents for more than 3,400 facilities, zones of impact from a catastrophe can extend from within 1 mile to 25 miles, based on the chemical, and more than a third of U.S. children live in such zones. Those worst-case plans are accessible to the public but can only be viewed in specialized reading rooms with a government monitor present — a secrecy regime justified by national security considerations that also functions as an information asymmetry between operators and the communities that bear the consequences of failure.

Deer Park, Texas, home to several major chemical plants, has experienced at least eight incidents since 2020 according to the CSB data — enough that the city trademarked a cartoon turtle called “Wally Wise Guy” to teach children how to shelter in place during a chemical emergency. Mayor Jerry Mouton told the Journal the town works “very hard with industry to make sure safety is a top priority.” The American Petroleum Institute stated that workplace safety is the oil and gas industry’s top priority and that industry initiatives have led to “significant declines in incidents over the past decade” based on its own survey data — a claim operating on a different dataset and different incident definitions than the CSB’s.

Systems Dynamics: Reinforcing Loops and a Weakening Correction

The structure contains two reinforcing loops working against safety and one delayed balancing loop currently being contested.

The deferred-maintenance loop. Aging equipment produces more frequent failures, which strain operating budgets and incentivize further maintenance deferral. This loop’s reinforcing quality holds when the operator is large enough to absorb accidents as operating costs — Kinder Morgan’s $70 billion scale can absorb 28 incidents — but the loop can break into a balancing correction for smaller operators where a single incident is existential, producing a bimodal distribution of risk governance across the sector. [Plant-level capital expenditure data correlated with incident history is not available from the substrate; whether the dominant behavior across the sector is absorption or correction remains unresolved.]

The coverage-gap loop. Limited regulatory coverage of high-hazard chemicals produces lower reporting, which weakens political pressure for oversight, which allows more incidents to occur outside the strictest rules. The EPA’s narrow dataset — excluding non-RMP facilities from its improvement narrative — reinforces this loop by making the problem appear smaller than it is.

The incident-to-correction loop. The chain from incident to investigation to regulatory tightening to reduced incident frequency is visible in the decline the EPA reports for RMP facilities over a prior period. For the broader universe of chemical hazards, that loop is slow, contested, and facing active reversal.

The dynamics exhibit features of what Peter Senge has described as “Shifting the Burden”: the core problem — aging infrastructure requiring replacement or significant refurbishment — is expensive to address directly. Deferred maintenance and regulatory restraint serve as symptomatic fixes that relieve immediate cost pressure while the underlying condition deteriorates. The burden of consequence shifts from the operator, which avoids capital expenditure, to the workforce, which bears physical risk, and surrounding communities, which bear evacuation, health, and property-value costs.

“Eroding Goals” is also visible: stated safety standards remain nominally intact while the operating conditions that made those standards achievable have degraded. The EPA’s position that facilities must “safely manage” hazardous substances regardless of RMP status describes a goal held in language while the enforcement mechanism to sustain it is proposed for elimination.

“Success to the Successful” operates at the enterprise level: large chemical and energy companies that have accumulated regulatory and political influence benefit from lighter oversight regimes, which reduce their compliance costs relative to competitors who might otherwise be compelled to invest in safety upgrades if rules were uniformly enforced.

The system’s response to stress is becoming more concave — more fragile — because two trends are operating simultaneously. The physical infrastructure is aging and its margins are thinning. The institutional safeguards are being weakened and politicized. The convergence is the structural fact. Fragility increases not because any single failure point has changed but because the distance between current operating conditions and catastrophic failure has narrowed on the physical side while the mechanisms that might detect and correct for the narrowing are being reduced on the institutional side.

The Recovery Window

The recovery window remains open. Reports of serious incidents exist. The structural causes are documented. Former officials from the agencies responsible have described the problem publicly. The question is whether the institutional will to act on that information can be sustained in an environment where the oversight body is proposed for defunding, the rules designed to address the gap are proposed for shelving, and the agency administering the most stringent safety program frames rising incident data as evidence of scope rather than substance. Without a change in the structure of incentives or the scope of transparency, the underlying fragility is likely to persist, and the next outlier event — a 900,000-gallon tank collapse, a dense urban evacuation — can be expected to arrive with little warning and a high bill.

Analytical techniques used in this piece

This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.

Fragility / Antifragility Audit
Asks whether a system gains or loses from volatility, shocks, and disorder (Taleb).
Stakeholder Mapping
Charts the parties to a situation — their interests, power, and alignments.
Systems Dynamics (Structural)
Maps a system’s structure — stocks, flows, and the architecture that shapes its behavior.
Bayesian Reasoning
Starting from base rates and updating beliefs proportionally as evidence arrives.