Not Even the EPA Can Say What's Polluting the Air

This story was originally published on ProPublica by Ava Kofman.

For decades, a factory on the outskirts of Portland, Oregon, has churned out hulking metal parts for Boeing’s commercial airplanes. Despite the steady pulse of its machinery, the plant maintains a low profile; Oregonians more readily associate Boeing with its historic headquarters up north in Seattle. Perhaps, I reasoned last spring, this helped explain why no one had noticed that the company’s satellite campus seemed to have unleashed an environmental catastrophe.

In 2016, Boeing reported to the Environmental Protection Agency that it had massively ramped up the amount of chromium compounds it was pumping into the skies of eastern Portland. For anyone who followed Erin Brockovich’s crusade against the dangerous chemical in Hinkley, California, this should have come as alarming news. Hexavalent chromium, as the highly toxic form of the metal is known, can cause lung, nasal and sinus cancers, trigger pulmonary congestion and abdominal pain, and damage the skin, eyes, kidneys and liver.

Although it is widely used in the aerospace industry to protect plane parts from corrosion, hexavalent chromium is such a potent carcinogen that in 2004 Boeing’s own environmental newsletter acknowledged that “it would be most desirable to eliminate the offending agent” altogether.

The only reason I’d heard about the issue at the Portland plant was because my colleagues Lylla Younes and Al Shaw, who have been reporting on toxic air since 2019, spent the past year calculating the cancer risks posed by industrial air polluters across the entire country.

Their first-of-its-kind analysis of the EPA’s modeled data initially found that Portland — despite its reputation as an eco-friendly, farm-to-table mecca — was home to one of the worst hot spots of cancer-causing air in the country, all because of a single plant’s emissions.

The clues to how this slow-motion disaster unfolded were buried in technical paperwork. In 2016, Boeing’s Portland plant told the EPA that it had released 1,954 pounds of chromium compounds into the air — a 1,000% increase from the year before, when it emitted 164 pounds.

In 2017, that amount soared to 5,556 pounds, where it hovered for the next two years. Because the factory makes planes, the EPA assumed that a fifth of the total chromium releases were hexavalent. When Al and Lylla crunched these numbers, they spotted a handful of elementary schools lying downwind of the factory. Our newsroom dispatched a team of reporters to investigate hot spots of cancer-causing air, and I was asked to look into Portland.

Everyone agreed that this was an urgent story, especially because no one else had reported on it. Although the data about these dangerous emissions was available to anyone who visited the EPA’s website, there had been no public outcry and no enforcement actions.

Between the appearance of flagrant corporate misconduct, the glaring regulatory failures and the staggering potential health effects, the matter seemed to call for just the kind of investigation that ProPublica was set up to pursue.

There was just one problem: None of it was true.

Much of the data that Americans rely on to understand pollution is self-reported by the polluters themselves. The EPA compiles this information in various catalogs, but the most ambitious of its efforts is the Toxics Release Inventory, which publishes land, water and air emissions data from more than 21,000 industrial facilities.

Landmark “right-to-know” legislation established the TRI program in 1986, and ever since it’s been celebrated for inaugurating a new era of environmental transparency.

In the beginning, then, our reporting team had little reason to suspect that a Fortune 500 company would submit numbers to a public database that would, incorrectly, make it seem like its plant had poisoned the skies of suburban Portland. It’s long been argued that information disclosure initiatives like TRI compel polluters to scale back immoderate emissions for fear that their names might otherwise end up on the front page of The New York Times.

If anything, we figured that corporations would be more likely to underreport their releases than to wildly overstate them.

It was with these assumptions that I first began to look into the Boeing case. But almost immediately, I hit a wall. Not one of the 39 people I interviewed, including workers at the Portland plant, had heard anything about the factory’s dangerous spike in hexavalent chromium emissions.

Nor did I find any circumstantial evidence to corroborate Boeing’s numbers in the thousands of pages of public records I requested — documents that included air permits, regional pollution studies, EPA files, city blueprints for the factory’s recent expansion, employee safety complaints and public school district emails.

Instead, I was uncovering information that appeared to directly contradict the federal database. Like the fact that chrome plating shops in factories across the country — including the Boeing Portland factory — emitted, on average, closer to a pound of chromium per year, according to a decade-old EPA survey of the industry.

Or, that in 2016, Boeing Portland told the Oregon Department of Environmental Quality that it had released a mere 0.53 pounds of hexavalent chromium — a figure that was several orders of magnitude lower than the 1,166.76 pounds of hex chrome that EPA records showed Boeing releasing the following year.

By now I’d started to feel as though the EPA data belonged to an alternate universe. So it was partly a relief to learn that Oregon’s regulators felt the same way. J.R. Giska, an air-quality engineer at the Oregon Department of Environmental Quality, told me that the state agency does not typically consult the federal database, since it has historically “found discrepancies in TRI reporting.”

Unlike the EPA, Oregon’s regulators require each facility to disclose in detail the underlying calculations it used to determine its emissions, and state officials rigorously check this work. The DEQ requests and reviews data from companies, Giska said, “at a level that the TRI reporting does not even come close to.”

Giska and his colleagues provided me with extensive documentation to support Boeing’s lower chromium figures, including granular calculations, stack tests and material safety data sheets — all of which five air quality scientists substantiated at my request.

Thanks to their time and expertise, I was able to see for myself that Boeing had erred. In 2016, the company started to account for all of the chromate-containing materials it had purchased while neglecting to account for the fact that most of those materials had been used on its planes or captured by its pollution-controlling equipment.

Up to that point, Boeing had refused to engage with me on the record about this riddle, but after I contacted the company a few more times, a spokesperson confirmed what I’d by then deduced independently. “Your reporting highlighted for us that we actually over-reported chromium emissions above what the site is emitting,” the statement said. “We will be submitting amended reports.”

As far as we were concerned, though, the case was far from closed. If the data was so off, how many other facilities had made similar mistakes? Why hadn’t the EPA investigated these emissions as a potential danger — or a potential outlier? And what was the purpose of TRI, ostensibly the most prominent pollution inventory on the planet, if its information was largely overlooked?

Our story was no longer that Boeing appeared to be poisoning Portland; it was that Boeing had said it was poisoning Portland four years in a row — and the EPA had ignored it.

Few facilities are actually measuring what’s coming out of their smokestacks; the EPA doesn’t require them to. Instead, many polluters approximate their releases using emissions factors — unreliable formulas that predict the average amounts a facility might expect to emit from a particular manufacturing process.

Most factors were developed decades ago and do not reflect the current industrial landscape. In 2006, the EPA’s Office of Inspector General concluded that emissions factors had “significantly underestimated” the pollution burden for several industries; some factories may be releasing twice as much pollution as the equation predicts.

Last fall, the EPA issued a warning that many emissions factors should “only be used as a last resort.” Even so, the agency essentially defers to companies to determine whether they’ve submitted “reasonable estimates” to the Toxics Release Inventory.

Todd Cloud, an environmental expert who’s consulted for both industry and environmental advocacy groups, told me that the data is often “80% math and 20% art — though some would call the art voodoo.”

The wide latitude for estimates and ambiguous standards for accuracy can create the impression that the data is “sort of presumptive,” said Craig Johnston, a former attorney at the EPA and a professor at Lewis and Clark Law School in Portland.

“I tell my law students all the time, ‘If you are uncomfortable with this dynamic and think it’s a fox-in-the-henhouse situation, the only solace I can offer you is that almost all of the relevant environmental data we depend on is generated by the companies themselves.’”

Although company officials must certify that they stand by the accuracy of their self-reported data, they’re allowed to revise their paperwork at any time. Earlier this year, The Intercept’s Sharon Lerner revealed that multiple facilities edited their TRI reports for the potent carcinogen ethylene oxide after the EPA determined that its cancer risk was 30 times higher than previously thought.

“With the submission of several online forms,” she wrote, “a huge portion of the country’s ethylene oxide problem disappeared — or at least the public record of it did.”

There are no studies on the overall reliability of the Toxics Release Inventory, but over the decades, researchers have documented concerns that industrial emissions may be severely underestimated. In 2004, for instance, environmental groups projected that releases of certain carcinogens were likely four to five times higher across the country than what had been reported to the TRI.

In 2019, three senators on the Environment and Public Works Committee requested that the EPA’s Office of Inspector General “assess whether data is being deliberately or unintentionally withheld from TRI reporting.”

In an interview over the summer, an EPA official told me that the agency continuously vets the data submitted to the TRI. It uses a program to spot forms that show vastly different emissions from the previous year or identical quantities multiple years in a row. Staffers also contact hundreds of facilities by phone and email, while the TRI Compliance and Enforcement program works to flag inaccurate submissions for extra scrutiny.

“We do our best to curate data to ensure that it is of the highest quality,” the official, who spoke on background at the EPA’s request, told me. “If the data has been inaccurate in the past, the facility has to update it.”

But who’s to say what’s inaccurate? Because TRI reports are largely based on estimates — and because those estimates are largely made in the absence of measurements — it’s incredibly difficult for the EPA to prove that a company has reported the wrong amount.

Nor would it be worth the gigantic effort, according to several former regulators, since the TRI program is not typically tied to the agency’s enforcement activities. When the EPA penalizes polluters for TRI violations, it’s almost always for neglecting to report their releases rather than for misreporting them.

After I spoke with the EPA in June, I asked about whether Boeing’s concerning chromium releases had ever been flagged internally. “Yes,” the agency replied. “When TRI staff contacted the facility regarding the accuracy of certain reported data, the facility told them the data were accurate.” The EPA apparently left it at that.

The agency’s laissez-faire approach calls the value of the TRI program into question, experts say.

“You don’t hear the phrase ‘thousands of pounds’ in the same sentence as ‘hex chrome.’ It’s like putting out plutonium,” said Cloud, one of the consultants who reviewed the Boeing documents at my request.

“Where there’s a story here is that this should have been immediately flagged by the EPA. Where are they? Are they asleep at the wheel? Why didn’t they catch this, or, if they knew it was wrong, ask the company to fix it?”

Shoddy data makes it difficult for the public to know what kind of pollution they’re being exposed to. It complicates the efforts of regulators — not to mention journalists — who try to identify the dangers that facilities pose to vulnerable communities.

It derails officials from investigating complaints. And it lowers the chances that anyone will step in to curb excessive emissions, which disproportionately burden communities of color.

People living near polluters often find themselves frustrated, and at times misled, by the gaps in the information at their disposal.

“All of these reporting requirements come out of the idea that a community should understand the risks posed by facilities nearby, but when you get down to the nitty gritty of it, there’s all of these off-ramps for the facilities,” said Mary Peveto, the founder of Neighbors for Clean Air, an advocacy group based in Portland, Oregon.

One of the most important things the EPA could do to improve its air-pollution data is require more facilities to actually monitor how much toxic air they release, according to all 29 environmental experts I interviewed about the issue.

When the Clean Air Act was passed in 1970, monitoring technology was in some cases too costly and complicated to be practical. Today, methods for measuring hazardous air pollutants are far more feasible, yet the practice remains rare, according to Scott Throwe, a former senior staffer in the EPA’s enforcement office.

Cloud, the environmental consultant, said, “We have machines you could put on a stack to make all this voodoo go away.”

The EPA’s new administrator, Michael S. Regan, told us that the agency plans to ramp up its monitoring capabilities with funds from President Joe Biden’s coronavirus stimulus package. An internal slideshow about the TRI program from 2016 acknowledges that “increased prevalence of personal pollution monitors could lead to questions [about] whether TRI reported data is consistent with ‘real’ levels.”

While monitoring is the best option, it’s not always the most practical one: Some devices are too expensive or insufficiently sensitive to be useful, and minuscule concentrations of certain air toxics can evade detection altogether. When monitoring isn’t possible or available, the EPA should shore up its standards for the estimates it accepts, Throwe said.

It could do this, for instance, by restricting the use of emission factors that it knows to be unreliable and by reconsidering how frequently facilities can alter their forms.

Federal officials might also do well to borrow from the playbooks of some of their successful state partners. After all, without the meticulous evidence compiled by the Oregon DEQ, we would not have been able to diagnose what was going on at Boeing Portland.

That state data exists thanks in part to activists like Peveto, who were so fed up with the EPA’s emissions inventory that they pushed state regulators to start their own database under a new law known as Cleaner Air Oregon.

“I didn’t see a lot of validity in trying to force EPA to improve or support their data,” Peveto said, “but I did see huge progress in having the Cleaner Air Oregon program provide us with stronger data we could believe in.”

The EPA said that it intends to release “more robust” air toxics information annually. “Good decision-making starts with good data,” its statement said.

“We have made major investments to improve detection technologies and emissions testing at industrial facilities. We also continuously work to improve our emissions inventories so that we can evaluate risks based on the most accurate information possible.”

By the summer, the seven of us who had been investigating toxic hot spots were fixated on what we’d taken to calling The Boeing Situation. If we had caught one major error that the EPA had overlooked, how many more were out there? We dropped everything to undertake our own vast data reckoning.

We weren’t starting from scratch. As far back as February 2020, Shaw and Younes had known that the data, like all data, had its limitations. Experts advised them to dig into companies whose emissions looked “fishy” or seemed unusually high for the facility’s size.

After they did this, four polluters said they realized they had made mistakes and revised their emissions forms. They were companies most people have never heard of. But once we stumbled across the Boeing problem, we wanted to know how many other large, sophisticated corporations might have also botched their data.

We attempted to contact the 200 facilities that appeared to be subjecting residents to the greatest cancer risk. We emailed, called, left messages and spoke on the phone with employees whenever possible.

In our correspondence, we asked companies to confirm the releases they reported to TRI for 2014-18, the years covered in our analysis, and noted that we had identified their emissions as elevating the estimated cancer risk in the area.

Of the 109 companies that responded, 71% confirmed that their reported emissions were correct, while 29% noted errors. When companies told us they’d erred, we asked them to resubmit their forms to the EPA, which most of them did, and we adjusted our model accordingly.

We also compared the federal data against state inventories and checked our work with dozens of regulators and scientists. (You can read more about how we went about all of this in our methodology).

In the end, some of the hot spots in our preliminary analysis shrank and others disappeared entirely. Since publishing our analysis last month, we’ve invited companies to let us know if they alter their TRI forms, and so far we’ve heard from three additional facilities that are revising their figures downward.

As journalists, we can’t help but wonder about the role that our questioning might have played in prompting some of these revisions. Perhaps other companies were contacted by the EPA, like Boeing, but only reexamined their forms following calls from reporters.

We also can’t rule out the possibility that some facilities might have revised their estimates downward because of the potential for negative publicity. It is not lost on us that it is far easier to get corrections out of companies that overreported their pollution than those that underreported.

Either way, the edits raise questions about how seriously companies take their duty to report to the federal government. I was struck by how many of the errors we encountered were caused by trivial oversights: Companies told us they had inserted a number in the wrong place, or employed the wrong method of calculation, or confused the amount of chemicals purchased for the amount they’d released into the air.

But one of the most common sources of error we encountered had less to do with the companies than with the way the EPA designed the program. The TRI paperwork doesn’t allow companies to distinguish between the forms of the heavy metal they’re releasing.

This poses a major issue for facilities that emit chromium compounds: Hexavalent chromium is a known carcinogen, whereas trivalent chromium is not, and yet the TRI directs companies to lump the two together. A handful of facilities told us that they didn’t release any hexavalent chromium, but because of the way TRI’s reporting system works, they couldn’t specify that.

The EPA makes this problem worse when it models the TRI data, since it assumes that some share of the total chromium compounds are hexavalent. (With Boeing, for example, the initial problem was caused by the facility grossly overstating its chromium releases, which in turn drove up the share that was assumed to be hexavalent.)

Internal memos and emails from the EPA reveal that it has long grappled with what it calls the “issue of heavy metals.” But the agency has yet to fix the issue, much to the frustration of current and former staffers.

“It’s a constant topic of conversation and a baffling problem for everyone,” an EPA employee familiar with the program told me. “I don’t think I’ve ever heard anyone at the EPA say it would not be useful and that they wouldn’t want to have it. It’s more, How can we go about it so that facilities will provide the estimates across the board?”

The fuzzy nature of the TRI data can give the agency cover not to act, said Wayne Davis, an environmental scientist at the EPA who worked in the TRI program until 2019. “If the agency can’t measure it,” he said, “it doesn’t have to manage it.”

Perhaps at this point it doesn’t go without saying: the Toxics Release Inventory, despite its flaws, remains the most comprehensive, readily available source of industrial air pollution data in the country.

When it comes to understanding what facilities are putting into the air, experts variously described TRI to me as “the only thing out there,” “the only game in town,” “the gold standard,” “a global leader,” “not the gospel but pretty good” and, on five separate occasions, “the best available data.”

Without TRI, many communities would remain in the dark about the health risks they face. “If you really want to try to understand where hot spots are, this is as good as it currently gets,” said Johnston, the law professor and former EPA staffer.

“It may or may not always be precise. It may or may not be accurate in every case. But it sure is a heck of a lot better than nothing. It should be a presumptive starting point for trying to understand aggregate impacts in these areas.”

That’s how we used it. Our map identified more than a thousand hot spots of cancer-causing air that warrant further investigation. It serves as a screening tool, one that allows residents and regulators to pinpoint places in need of greater air-monitoring efforts, data-quality inspections and enforcement activities.

If anything, experts say, the risks we show are typically an underestimate: the floor, not the ceiling. Our team — now along with dozens of newsrooms — has been reporting on how communities in these hot spots are affected by the factories in their midst.

Since we published our analysis, we’ve heard from over a thousand people living in or near more than a hundred toxic hot spots in 34 states. They’ve circulated petitions, packed town halls, applied for air-monitoring grants and met with environmental regulators.

The EPA is paying attention, too. On a recent tour through some of the most toxic hot spots in the country, Administrator Regan told me, “We’ve looked very carefully at your reporting and we’re incorporating much of it into our refined and revised system ourselves as well, so that we can begin to address these issues.”

Community leaders and environmental advocates have stressed that the limitations of the data should not delay the agency from taking action to curb dangerous emissions. “The data show a problem that EPA needs to address,” said Emma Cheuse, an attorney and air toxics expert at the advocacy group Earthjustice. “You can constantly improve the data, and more research will always be helpful, but the EPA also needs to act on the information we have now.”

Feature Photo by Maxim Tolchinskiy on Unsplash