Deregulatory Disclosure Regimes and Masked Absences in Fracking Chemical Data

Abstract

Vocal opposition to fracking (the practice of fracturing the subsurface to increase oil and gas production), from shale deposits in Pennsylvania to oil fields in California, has led to increased disclosure about chemicals in fracking fluid. This paper questions the conflation of data abundance with meaningful action or response, drawing on the authors’ collective years of experience working with FracFocus, the most comprehensive publicly available fracking chemical database in the United States. We identify and characterize a range of absences and ambiguities in FracFocus, the most obvious of which are trade secrets, which are explicitly marked absences. Pervasive structural and systemic absences are harder to see but constrain the usefulness of FracFocus data. Using perspectives from environmental STS and critical data studies on deregulatory knowledge production, we analyze these absences to think differently about disclosure. Rather than trying to “fill in” these holes, we use their form to better understand the ways in which regulatory structures shape knowledge production, as well as the much more profound questions forced by the unruly nature of chemical behavior. If investments in disclosure-as-governance rely on longstanding ideas of isolated and discrete chemicals, these holes emphasize the importance of reckoning with chemical mutability and uncontrollability.

Introduction

“Those trade secret chemicals, they’re just a black hole, aren’t they?”

—Open-access fracking data workshop participant, 2024

“We have heard our communities and concerned residents asking for data related to human health, and we have seen them left confused and empty-handed from the results of flawed studies and inflammatory media headlines. We stand ready to respond.”

—CNX Resources Corporation, 2023

These two statements both address the disclosure of chemicals used in fracking, but from two very different positions. (Fracking is shorthand for “hydraulic fracturing,” which sends pressurized liquid underground to fracture rock and increase access to pockets of gas or oil.) The first comment, from a 2024 workshop with environmental activists working at the local level across the United States, epitomized common frustrations around trade secret chemicals within existing data disclosure mechanisms. The second is from the “radical transparency” campaign of CNX, a natural gas company based in Pittsburgh, marketed as a route to “mutual trust” between corporations, communities, and regulators. The two speakers would likely see themselves on opposite sides of most debates about the health and environmental impacts of fossil fuel economies. Yet, just a few months apart, both emphasized lack of knowledge as a primary issue, and transparency through data as a solution. By juxtaposing shortcomings in existing disclosure mechanisms alongside the ease with which industry can mobilize languages of transparency and disclosure, these two statements demonstrate insufficiencies in the “disclosure” frame itself.

This article investigates the recent and increasing profusion of data on fracking to question the conflation of data abundance with meaningful action within the context of US regulatory and disclosure regimes. We are a trio of STS and big data scholars, each of whom has studied big data and the (de)regulatory science of oil and gas in different ways. Drawing from scholarship on environmental data justice and absences in scientific knowledge production, we demonstrate a range of absences and ambiguities that remain in, and constrain the usefulness of, this abundance of disclosed data. We focus on FracFocus, which is the largest public national dataset in the US on fracking chemical use. FracFocus data are widely used by scholars (including ourselves), regulators, and industry employees; they also underpin CNX's radical transparency proposals. However, as we explain below, we estimate that no FracFocus disclosures are complete: all contain some absence or ambiguity, which we call “holes.” Beyond a desire for or fantasy of full disclosure, what can the shape of these holes tell us about the structures of environmental governance within which fracking chemicals have emerged as a notably unregulated gap?

Here, we expand an index created by co-author Gary Allison of the many kinds of absences in FracFocus. This work established what one might imagine these disclosures to offer—what we call the “naive expectation” of disclosure—and then methodically showed the many ways in which holes, both invisible and visible, undermine that original expectation. We expand that analysis by attending to the function of deregulatory science (Mansfield 2021), or how scientific knowledge can be mobilized to weaken protective regulations, as well as the insufficiencies of data disclosures as a governance mechanism.

In that sense this article both experiments with data ethnography and records our learning process, because each of us began with some version of that naive expectation; through our work together we came to understand these various holes and their significance. We begin by contextualizing fracking chemicals within larger landscapes of fossil fuels and deregulatory science, and then describe our work with fracking chemical data, including an introduction to FracFocus and our open-data interface with FracFocus, called Open-FF. We situate our work within growing scholarly attention to ignorance, uncertainty, and disclosure regimes in environmental governance. Finally, we provide an index of the kinds of absences that remain in FracFocus and what they tell us about environmental regulation more broadly.

While the aim of our analysis is not to “fill in” holes in fracking data, this does not mean that disclosure is not important: one of our goals is to demonstrate how to use even poor data to learn about landscapes of toxicity. Yet, too often, critiques of data “result in a call for more or better data on which to act… [and this] has become part of regulatory inaction” (Boudia et al. 2021, 5). Aligned with Boudia et al.’ focus on the spatial and temporal residues of chemicals and the apprehension of them, we aim to document and explore the shape of holes in FracFocus to better understand the structures of their production and their productive effects: what kinds of futures do they propel, and what do they foreclose?

Ultimately, we find that the holes within chemical disclosure databases teach us about the insufficiencies of governance-by-disclosure, as well as the impacts of concomitant frames that assume chemicals to be isolable and discrete entities whose behavior can be predicted, constrained, and contained. Although for many it might be easier to imagine fracking chemicals and their behavior as known but simply not disclosed in an example of corporate malfeasance, these absences require a much deeper kind of relearning for many environmentally minded academics, industry professionals, and local residents alike. If investments in disclosure-as-governance rely on longstanding ideas of discrete and controllable chemicals, these holes emphasize the importance of reckoning with chemical mutability and uncontrollability.

Fracking Chemicals in the Landscape of Fossil Fuel Hazards and Deregulatory Regimes

Fracking chemicals are only one part of a much larger landscape of socio-ecological hazards associated with fossil fuels and deregulatory regimes that at best fail to control these hazards, and at worst foster them. Here we situate fracking chemicals in the landscape of fossil fuel reliance, fracking as a production process, and the suite of deregulatory actions at every step.

Fossil fuels lie “at the heart of the interconnected crises we face, including climate change, racial injustice, and public health” (Donaghy et al. 2023, 1; see also Appel, Mason, and Watts 2015; Bond 2022; Daggett 2019). Yet regulation of fossil fuels and their effects is notably lax, particularly in our US context but more broadly as well. While some regulations have led to some improvements (like the 1963 Clean Air Act which reduced smog and soot in many US cities), these successes are partial and hard-won, and loopholes both formal and informal abound (Cupas 2008; Warner and Shapiro 2013; Wiseman 2009). Lack of regulation, the difficulty of implementing already late and minimal regulation, and judicial undercutting of environmental regulation lead many to see the US federal regulatory system itself as part of the problem: structured to allow pollution, minimize disruption to polluters, and facilitate capital accumulation rather than protect public and ecological health (Liboiron 2021; Kojola and McMillan Lequieu 2020; Murphy 2006).

One major set of concerns is about how fracking props up fossil fuels and their associated problems, from climate change to petrochemical exposure. In the context of climate change, proponents tout fracking (especially for natural gas) as providing a bridge away from coal, while critics note instead that these new technologies expand reliance on fossil fuels by opening further reservoirs of oil and gas that would otherwise not be accessible (Deberdt and Le Billon 2024; Willow and Wylie 2014). Moreover, oil and gas extraction and transport themselves release enormous quantities of methane, an especially potent greenhouse gas, through both purposeful venting and unintentional leaks known as “fugitive” emissions (Heede 2014; Mar et al. 2022). In addition, oil and gas are fuels as well as feedstock for petrochemicals, including plastics, solvents, and pesticides. Acute and chronic exposure to petrochemicals, at various stages of their lifecycle from production to disposal, is linked to a wide range of health problems (Davies 2022; Donaghy et al. 2023; Willis and Buonocore 2023). The current regulatory system's inability to adequately protect human and ecological health from these hazards is well documented (Aker et al. 2024; Deziel et al. 2022; Hatzenbuhler and Centner 2012; Mansfield 2022), and yet fracking is expected to increase with recent investments in hydrogen fuels and liquid natural gas hubs as part of the US energy transition. ¹

Another set of concerns is about the specific hazards related to fracking as a set of production methods. Studies have correlated living near fracking sites with a range of health problems, including cardiovascular and respiratory disease, stroke, childhood leukemia, and adverse birth outcomes (Elliott et al. 2017; Hu et al. 2022; McAlexander et al. 2020; Tran et al. 2021; Trickey, Chen, and Sanghavi 2023). Such studies cannot tease out the precise causal links between fracking and poor health outcomes, but other studies have identified a range of potential factors, including noise (Richburg and Slagley 2019); methane, which can affect both health and climate (Alvarez et al. 2018; Mar et al. 2022); and a range of other hazardous pollutants released alongside greenhouse gases, like toxic VOCs (volatile organic compounds) and soot (fine particulate matter) (Banan and Gernand 2021; Li et al. 2022; Purvis et al. 2019; Zhang et al. 2023). Although never listed as an individual's cause of death, particulate matter is a leading cause of mortality worldwide because it causes serious health problems such as heart disease (Michikawa et al. 2022; Peterson et al. 2020; Ward-Caviness et al. 2020). And while the denial of climate change science catches the public imagination, the deregulatory science of particulate matter has played a quieter but foundational role in decades of failure to adequately protect public health and address environmental injustices (Dillon and Sze 2016; Mansfield 2022; Mikati et al. 2018; Tessum et al. 2021).

The chemicals used during fracking are an area of special concern within this broader landscape. These ingredients, many of them petrochemicals made from fossil fuels, can do anything from lowering the fluid's friction to creating acidic conditions to dissolve rock (Chittick and Srebotnjak 2017; Finšgar and Jackson 2014; Hill, Yadav, and Khan 2022; Stringfellow et al. 2017). Hundreds of chemicals have been documented in fracking operations, some in vast quantities: those used in the largest volumes can exceed hundreds of thousands of pounds per frack (Allison and Underhill 2024; US Department of Energy 2014). Fracking proponents claim that these chemicals are safe, yet our prior analysis showed that many are chemicals of concern under the 1972 Clean Water Act and/or 1974 Safe Drinking Water Act, though when used in fracking, they are exempt from regulation under such laws due to the “Halliburton loophole” discussed below (Underhill et al. 2022, 2023).

As our introduction suggests, critics cite both known chemicals of concern and holes in knowledge as reasons for worry, especially given that the nature of fracking (horizontal drilling, inducing fractures, and chemical dissolution) means there is potential for these chemicals to spread and persist far beyond the original wells (McMahon et al. 2017; Osborn et al. 2011). There is also the question of what to do with the wastewater (rebranded as “produced water”) after it comes back to the surface of a fracking operation. Wastewater contains the chemicals added to fracking fluid and many more—including benzene and other aromatic compounds, heavy metals, and radioactive isotopes—that it picks up “naturally” underground (Luek and Gonsior 2017; McDevitt et al. 2019; Nobel 2024).

While our analysis is focused on a US context, similar “regulatory gaps” have been described in the UK (Hawkins 2015) and across the EU, even within a more cautious regulatory setting (e.g., Bomberg 2017; Goldthau and Sovacool 2016). Similar “regulatory lags” (Maloney 2015) have been described in Australia, where the federal government cannot compel states to require the disclosure of chemicals (Esterhuyse, Vermeulen, and Glazewski 2019), in part because—like Canada—it devolves much jurisdiction over fracking to provincial or state regulators (Ingelson and Hunter 2014; Carter and Eaton, 2016).

Thus, the question of fracking chemicals is more expansive temporally and spatially than the fracking operation itself: it is about the exposure of workers and nearby communities, about effects to soil and groundwater when some fraction of the chemicals that go down the frack fail to return, and about the production of these chemicals in the first place. It is also about the billions of gallons of contaminated water that result: some reused in fracking (see below), some pumped back underground as waste (often after being trucked to distant injection sites), and some, in a masterful public relations move, “recycled:” spread on roads to suppress dust and melt ice or used to irrigate crops (Cooper et al. 2022; Miller et al. 2019; Zhang et al. 2018). Fracking chemicals may be one small part of the fracking problem, which itself is just one aspect of fossil fuel and petrochemical problems, yet this one “small” problem is itself enormous.

FracFocus and Open-FF

This suite of concerns inspired our interest in fracking chemical disclosure, though we came to it from different relations and commitments. Authors Vivian Underhill and Becky Mansfield are social scientists and STS scholars focused on environmental governance and environmental justice; Author Gary Allison is a data scientist. Vivian spent many years doing ethnographic and collaborative work with environmental justice groups organizing around fracking in California's San Joaquin Valley. Located in Ohio and attuned to issues of fracking chemicals and waste as a local and state issue, Gary has been curating and consolidating marginal industry data sets to make them more accessible. Becky, also in Ohio, comes to the question of disclosure through academic research on “chemical geographies” and the political ecology of regulatory science.

Our work together is also grounded in digital communities that coalesced around the function and circulation of national environmental data. We began working together through the outward ripples and networks catalyzed by groups like the Environmental Data Governance Initiative (EDGI). EDGI originally coalesced in 2016 to document changes to environmental data and regulation under the first Trump administration; as we were putting the finishing touches on this manuscript, the second Trump administration took office and redoubled the importance of archiving and monitoring public environmental data. However, this project is also situated within a broader, longstanding community of data scientists, researchers, and data justice advocates investigating the insufficiencies of right-to-know legislation and of data as a regulatory form that include, but are not limited to, the deletion, removal, or obfuscation of data.

In 2019, Gary developed an open-source system (Allison, 2024b) which extracts and analyzes data from FracFocus, a third-party disclosure system. FracFocus was initiated in 2011 during a period of intensifying pressure to regulate and disclose the chemicals used in fracking. Supporters of fracking claimed the chemicals were benign and therefore not worthy of disclosure, but the fracking industry's desire for secrecy about these chemicals allowed for speculation about worst-case scenarios. This speculation proved successful for anti-fracking activism, as even industry supporters acknowledged (Kinchy and Schaffer 2018). The CNX radical transparency campaign seems to echo this experience by framing transparency as demystifying fracking chemical use and therefore placating public concerns. In this context, FracFocus as an industry-led disclosure mechanism serves a public-relations function as much as a data-gathering one. FracFocus describes itself as “a one-stop, easy-to-understand public resource for consumers” (FracFocus.org n.d.). Some scholars have made productive use of FracFocus data (Hill, Yadav, and Khan 2021; Trickey, Hadjimichael, and Sanghavi 2020), but critics call FracFocus “transparency on [the industry's] own terms” (Kinchy and Schaffer 2018, 1018) and “opaque transparency” (Avidan, Etzion, and Gehman 2019, 198).

Operators submit chemical disclosure forms to FracFocus, which then indexes and publishes the raw disclosures in PDF format for bulk download. As of April 2024, they had information for over 204,000 unique wells and over 212,000 unique fracking jobs reported by 1,600 individual companies (FracFocus.org 2024). Within a given disclosure, anywhere from 10 to 100 individual chemicals might be disclosed; we refer to each chemical within a given disclosure as a “record.” From 2011 to 2025, this amounts to 6.4 million records and counting (Allison 2025). Most disclosures also include the percentage of the disclosure that a given chemical represents, the amount of water used, and the trade-named products of which each chemical is a part. This is clearly a situation of data abundance: even with the statistics above, it is difficult to communicate just how much total data FracFocus holds.

However, FracFocus may be one-stop but it is not so easy for consumers to understand. It was introduced as a voluntary measure, is difficult to use, and although most oil and gas states now require reporting to FracFocus, there remains no federal-level disclosure or monitoring requirement (Konschnik, Holden, and Shasteen 2013; Tansey 2018; Trickey, Hadjimichael, and Sanghavi 2020). For example, on the FracFocus home page, users are confronted with a choice of viewing single PDF disclosures or downloading all 6 million+ records; they will find chemicals identified by confusing or conflicting labels; company and product names are not standardized, making searching difficult; and there is little usable information about the actual quantity of the chemicals. Compiling just a list of the chemicals a company uses in one's county can be a monumental task.

It is in this context that Gary developed Open-FF as a vehicle to make these data more accessible and understandable. While it is tempting to write off FracFocus data altogether as fatally flawed industry data, the aim was to simultaneously provide avenues for critique and make the data more useful for understanding which chemicals are used, in what quantities, where, and the known hazards associated with them. Open-FF includes multiple analyses and data-analysis tools coded by Gary in extensive conversation with the other two authors. See Underhill et al. (2023) and related Github documentation (Allison 2022) for discussion of Open-FF's data curation and quality-control methods.

While Open-FF was itself inspired by evident problems with FracFocus, it was in the process of developing and using Open-FF that we most fully understood the significance of the absences and ambiguities in this disclosure regime. In our first project together, Vivian and Gary used Open-FF to quantify the outcomes of the Halliburton Loophole (Underhill et al. 2023). This loophole refers to the 2005 National Energy Policy Act's exemption of fracking activity from the Safe Drinking Water Act. It is nicknamed as such to call attention to then-US Vice President and former Halliburton CEO Dick Cheney's instrumental role in its formulation and legislative passage. As we calculated the total mass of chemicals used in fracking that would otherwise have been regulated under the Safe Drinking Water Act, the consistent absences in FracFocus's data were striking. The most obvious of these absences was the use of trade secret protections to avoid reporting specific chemical identities. Therefore, using FracFocus data to investigate FracFocus's own insufficiencies, we next published an analysis of the large (and increasing) use of trade secrets in fracking chemicals (Underhill et al. 2024).

Building from these investigations, in 2023 and 2024 Vivian and Gary ran a series of community-specific workshops on understanding what can be learned about fracking chemicals using Open-FF. Environmental groups in California, Colorado, Oregon, Pennsylvania, and Texas participated, as well as artists and other scholars. Conducted over Zoom, participants came with specific experiences with fracking in their watersheds; we shared how the data work and what tools might be useful in their contexts, showing how to link an aerial view of fracking sites to the lists of chemicals disclosed (and not) for those specific wells. In every workshop, the question of what we didn’t know had a gravity of its own: the “black holes” epigraph came from one of many discussions about the problems posed by these absences.

This inspired us to create a larger index of ambiguities, absences, and unknowns in FracFocus data. Trade secrets turn out to be a relatively “friendly” or accessible kind of unknown: they announce themselves, are clearly recorded in disclosure forms and in the aggregate data, and often come with the mass used of the withheld chemical. Trade secrets mark themselves absent – butthe many other holes cannot be so easily measured.

As we were grappling with these other holes, Pennsylvania and CNX announced their new transparency campaign, which the Governor's office lauded as a “historic collaboration” between state regulators and natural gas producers – one that would, in CNX's words, “definitively confirm for all stakeholders that there are no human health problems related to natural gas development” (CNX Resources Corporation 2024). However, this “historic collaboration” provides little new, as it still revolves around public disclosure of chemicals used in drilling and fracturing. For fracking, CNX simply reformats FracFocus disclosures (and all their insufficiencies) and adds the CNX logo. The nonprofit FracTracker Alliance, which provides data and maps about hydrocarbon extraction, incorporated our index of absences in their public response to the CNX Radical Transparency campaign (Jones 2024). (FracTracker now sponsors Open-FF and publishes the index of holes on their website (Allison and Underhill 2024)). Ongoing investigation of the data held in FracFocus will continue to be necessary as its reach has recently expanded to include non-oil and gas wells, including those used for carbon sequestration, critical and rare earth mineral extraction, and other technologies of the emerging decarbonized energy landscape.

Disclosure and Transparency as Environmental Governance

Both FracFocus and CNX's Radical Transparency campaign function within a logic that conflates data abundance with good environmental governance. This conflation is what we want to disrupt: FracFocus may provide an abundance of data, but it also produces absences and unknowns not despite but through that abundance. Not only must scholars approach disclosures with a critical eye to their conditions of production, but, we assert here, the absences in these data are a powerful avenue to better understand the structures of power in which data is disclosed, used, and given meaning (e.g., Agostinho et al., 2019). Similar questions of disclosure and trade secrets pervade the US chemical regulatory regime more widely.

As a large body of science studies scholarship demonstrates, intentional corporate secrecy and information withholding are just part of the dynamics of unknowing about chemicals and their hazards (Boullier and Henry 2022; Henry et al. 2021; Kroepsch and Clifford 2022; Proctor and Schiebinger 2008). Attempts to easily distinguish between “good” (real, objective, revealing) and “bad” (made-up, corporate, obfuscatory) science obscure the longstanding STS insight that all scientific work is produced through a range of always already social and political practices (e.g., Haraway 1988). Against a simple binary, we know that mainstream and accepted scientific knowledge is and has been central in formulating and maintaining social and environmental harms and injustices (Gould 1996; Harding 1986; Stepan 1986). On the flip side, the many attempts to dismiss evidence of these harms and injustices are not best understood as non-science but as deregulatory science – that is, the mechanisms through which the state produces and mobilizes scientific knowledge against protective regulations (Mansfield 2021).

Ignorance and knowledge are often co-produced through regulatory processes. Murphy's (2006) well-known frame of “regimes of imperceptibility” suggests, through a detailed ethnographic study of scientific work at the EPA, that regulatory science consistently produces some things as imperceptible while making other results perceptible. Similarly, “undone science” is “a kind of non-knowledge that is systematically produced through the unequal distribution of power in society” (Hess 2015, 142), while others have shown how EPA actions simultaneously create knowledge and regulatory ignorance (e.g., Frickel and Edwards 2014). Indeed, unknowing can be propagated through the sorts of datasets, measurements, and monitoring apparatuses that appear as good science and that activists often demand (Calvillo 2018; Marzecová and Husberg 2022; Nafus 2024; Tadaki 2024).

If “disclosure” and “(radical) transparency” appear to be the obvious way of addressing whether fracking is harmful to human and environmental health, this is because disclosure has become part of a “common sense” around how publics relate to knowledge. In the US in the 1970s, the Toxic Substances Control Act (TSCA) authorized EPA to protect public and environmental health by making chemical toxicity data public. But the law itself, and its implementation under Reagan, continued to allow non-disclosure of “confidential business information” (Creager 2021; Hepler-Smith 2019) and undermined the usefulness of toxicity data even when disclosed. Subsequent demand for greater disclosure responded to frustrations over this non-disclosure (Cranor 2017; O’Reilly 2010) and became especially prominent after the 1984 Bhopal disaster in India, in which the accidental release of methyl isocyanate gas led to the deaths of over 3,800 people and long-term health complications and premature deaths for many more (Broughton 2005; Fortun 2001). The US passed several new chemical disclosure laws, including the 1986 Emergency Planning and Community Right-to-Know Act, which created the Toxics Release Inventory. (The oil and gas industry has been excluded from their reporting requirements from the beginning.)

By now, scholars have well articulated how disclosure has become a governance structure of its own rather than one part of an overall strategy: an end rather than a means. This “governance by disclosure” (Gupta 2008) poses individual knowledge as an a priori “good” that will lead to improved regulatory frameworks and industry practices. As such, it is compatible with neoliberal forms of governance that emphasize not only voluntary options for industry but individual (consumer) choice. A prominent example is an overemphasis on labeling (organic, BPA-free, fair-trade, non-GMO, etc.) at the expense of meaningful regulation (Guthman 2007). Though disclosure might masquerade as a commonsense policy, it becomes a “targeted transparency” that enrolls ideas of “individual choice, market forces, and participatory democracy”—often to depoliticizing effect, while allowing light government action (Fung, Graham, and Weil 2007, 5; Kinchy and Schaffer 2018).

Yet the connection between information and improved governance is by no means obvious. As Fortun (2004) emphasizes, environmental information systems also encode particular forms of subjectivity and assumptions of what knowledge does. Because of their “belief in information as the sine qua non of transparency,” disclosure regimes can lead to the fetishization of information as a commodity, an ideal, and a deeply overdetermined concept (Adams 2018; Hansen, Christensen, and Flyverbom 2015, 117). Data can defer more substantive action (Horgan forthcoming), and data transparency can be weaponized to stall regulation by rejecting scientific studies argued to be inadequately transparent (Mansfield 2021, 2022; Underhill et al. 2017). Meanwhile, corporations from oil and gas to mining have increasingly embraced performances of transparency and accordance with regulation to assert their legitimacy and social responsibility (Kojola and McMillan Lequieu 2020).

Finally, the very act of disclosure can obfuscate what is not being disclosed: the gaps and silences. Especially in the case of very large data sets—such as FracFocus—users may not only experience “drowning in disclosure” (Gupta 2008), but the massive datasets in which one might “drown” produce the impression of completeness and objectivity. As scholars of data have shown, this faith in the objectivity of big data is misplaced; all data are political (i.e., shaped by power relations) and produced through various social practices and relations (Gitelman 2013; Nost 2022).

To better understand the shape of FracFocus's absences and understand the working of disclosure as an operation of power, we use FracFocus to explore its own absences. First, we join other critical scholars of environmental data to not shun such data but work with what they can tell us, using these data in novel ways to generate useful information. For example, HazMatMapper is an online tool for exploring hazardous waste trade between the US, Mexico and Canada using EPA data obtained through Freedom of Information Act requests (Nost et al. 2017; Rosenfeld et al. 2018). The Environmental Enforcement Watch similarly uses faulty and incomplete EPA data to track and monitor facilities’ environmental violations. Horgan et al. (2023) drew generative results from EPA's EJScreen even while critiquing its methods. The development of Open-FF is in this vein, with its goal to make FracFocus data more accessible and usable. Open-FF, and analyses that use it, are examples of what imperfect data can still tell us about the chemical footprint of the petrochemical regime. Second, we use Open-FF's interface with FracFocus to examine the shape and pattern of the holes in the data themselves and use the shapes of these absences to better articulate the limits of “governance by disclosure” and, more broadly, the limits of premising environmental protection on better information.

Using Absences in Data to Understand Data's Absences

From the detail and abundance in FracFocus disclosures, one might reasonably expect that FracFocus as a chemical disclosure instrument documents all the chemicals used in a fracking operation. We might expect something like the schematic in Figure 1, in which what goes into the ground during fracking is directly represented in the FracFocus disclosure.

Figure 1.

The Naive Expectation of Disclosure. The “naive expectation of disclosure” is that there is a one-to-one correspondence between the chemicals that go into the ground in an individual frack and the chemicals listed on the disclosure form. The left arrow is a schematic of all the chemicals that might go into the ground, with abstract names like water” or “sand.” The right arrow shows what the disclosure would show: in this “naive expectation,” they are exactly the same. Created 2023 by Gary Allison.

We began to call this the “naive expectation” of disclosure. This expectation includes the anticipation that the identity and quantity of the chemicals used would be clearly stated: in Figure 1, what goes into the ground in a frack is directly and completely reported in a disclosure form. But how many disclosures actually meet this “naive expectation”? Our estimate is zero. That is, we estimate that all disclosures contain one sort of absence or another.

Resisting the pull to just “fill in” these data holes, the more interesting question becomes: what can these absences tell us about how regulatory and disclosure regimes function? In this section we parse the types of absences we have encountered to explore the shape of these holes as themselves a kind of chemical “residue” (Boudia et al. 2021). We move from more “visible” to more “invisible” holes that require specialized knowledge to notice, and even with specialized knowledge, that are almost impossible to rectify.

Trade Secrets and Ambiguous Data

Perhaps the friendliest holes in FracFocus are those that announce themselves (Figure 2A). We call these visible holes, of which the most common form is proprietary claims or trade secrets. These designations, in which a given operator or supplier reports that a chemical was used (and often the mass used) but does not report the identity of the chemical, are allowed at the federal level and by most states that have state-level fracking regulations. ² More than 80 percent of disclosures report at least one proprietary record and more than one quarter of disclosures hide at least 25 percent of the chemicals with these trade secret designations. The cumulative mass of these records from 2014 to 2022 is over 10 billion pounds (Underhill et al. 2024). In Figure 2A, these are indicated as “trade secret” in the disclosure column.

Figure 2.

Holes in FracFocus Disclosures, Based on Analysis in Open-FF. Represented here are non-disclosure of: (A) Proprietary chemicals and chemicals with ambiguous identification (top left). (B) Chemicals not listed on Safety Data Sheets, which are produced only for occupational health and safety and do not include all potentially hazardous chemicals (top right). (C) Chemicals in water, even when using fracking wastewater for additional fracks (bottom left). (D) Chemicals within UVCBs (“Unknown or Variable Composition, Complex Reaction Products or Biological Materials”) (bottom right); the provided example is heavy aromatic solvent naphtha, which can contain the individual chemicals listed here (and many more). Created 2023 by Gary Allison.

Another semi-visible hole is ambiguous data: these are records that are not labeled or are labeled in a way that cannot be resolved to a specific chemical. For example, instead of reporting a concrete chemical identity, a dummy value (e.g., “xxxxxx-xx-x”) is used instead, or the CAS number (short for Chemical Abstracts Service, which gives a unique number to every known chemical compound) might be blank. In Figure 2A, these are the boxes labeled “Missing ID.” Although these records appear in a disclosure, they do not include enough information to know their identity, and therefore their potential hazards. At least one ambiguous chemical record is present in 23 percent of disclosures in FracFocus (Allison 2024d).

Related to these ambiguities, FracFocus contains obvious errors that go uncorrected. For example, although water and sand typically make up roughly 88 and 11 percent of the fracking fluid respectively, a substantial number of disclosures report sand percentages greater than 90 percent with small percentages of water—suggesting the numbers are switched (Allison 2024a, 2024c). Many of these errors have existed in FracFocus since 2015, and new disclosures with this pattern still regularly appear. Such large discrepancies undermine the integrity of entire disclosures. To date, there are over 2,000 disclosures, spread over 20 states and hundreds of companies, with sand values over 50 percent. The frequency and magnitude of ambiguous and erroneous data shows that, indeed, disclosure has become an issue of “checking the box,” and many regulators and industry alike consider it sufficient to simply fill out the form, while passing on the responsibility for the quality of the data disclosed on the form.

Partial Lists

Less visible holes arise in the case of chemicals injected in fracking operations that aren’t mentioned at all in FracFocus disclosures, so that we only know they exist from evidence outside of FracFocus. This often occurs through partial ingredient lists, which are indicative of the wide latitude manufacturers are given in deciding what is “hazardous” and what is not—and, therefore, what is reported and not. These are represented in Figure 2B: the “nonhazardous” chemicals 5 and 7 and the “inert” chemical 6 do not show up at all on the disclosure form. One prominent mechanism here is reliance on the lists of chemicals included on Safety Data Sheets (SDS). The Occupational Safety and Health Administration (OSHA) requires SDS for many trade-named chemicals, to help protect worker health and safety and to inform emergency response teams in the event of spills or injuries. But SDS were not designed to be an exhaustive ingredient list and they do not include all the chemicals in a particular product: they focus on ingredients that are potentially hazardous in a workplace setting. Still, many companies disclosing to FracFocus exclusively use SDS lists to streamline their disclosures, thus limiting what they report. (There is space for operators to add other chemicals, and this could include the non-SDS chemicals of a product, but that is not explicitly required.)

This creates a major absence in data because what may be considered nonhazardous in an occupational setting may be hazardous in a larger environmental setting. One example is PFAS, so-called “forever chemicals.” Dangerous in minute quantities in water systems, they are also known to be highly effective surfactants in fracking. Although EPA is now working on a response to their widespread use, they have been considered inert or at least nonhazardous in most products, and so may have been omitted from SDS. While FracFocus contains some records of PFAS chemicals, the common practice of only relying on SDS sheets is likely hiding more widespread use. Precisely because of their invisible nature, these invisible non-SDS chemicals are hard to quantify. Among disclosures for which we have enough information to calculate, over 250,000 uses of trade-named products report less than 50 percent of component chemicals (Allison 2024c).

This set of holes highlights how slow the process of studying chemicals is for EPA and other environmental regulatory agencies, and how far ahead chemical engineering is from chemical regulation. Some of the issues around deeming chemicals “nonhazardous” might be allayed if regulatory science were better able to keep up with experimentation in their use and application. For instance, EPA has only defined legally binding maximum contaminant levels (MCLs) for 79 chemicals (US EPA 2015a, 2015b). Similarly, a 2016 study found that only 8 percent of the 1,076 chemicals reported in hydraulic fracturing fluid had a chronic oral reference value available (Yost et al. 2016). ³ Yet there are hundreds of thousands of chemicals in production and use, with more being developed every year. Even the process to nominate a particular chemical for further study by the EPA is long, circuitous, and often deeply political (Mansfield 2021). This dovetails with a regulatory system in which chemicals are presumed “innocent until proven guilty,” unregulated until hazard is proven.

This leads to a kind of chicken-and-egg problem. Chemicals that are presumed nonhazardous are less likely to be reported in FracFocus, especially if operators are relying on SDS sheets. But if scientists don’t know they are present at all, it makes it much harder to study their actual behaviors and effects. This, then, goes one step beyond the prohibitive nature of the sheer number of chemicals requiring research. It makes it more difficult to provide “proof” of hazard, let alone calculate harm or risk.

Produced Water and “Unknown and Variable Compositions”

Water use represents another invisible hole in the fracking chemical disclosure data (Figure 2C). For most fracking disclosures, water is the primary carrier fluid, representing over 80 percent of the fracking fluid. Water can come from a range of sources: nearby surface water, groundwater, or even municipal water sources. But it can also be fracking wastewater, reused by operators. The composition of this wastewater, though hard to know for certain, often includes a combination of “natural” materials acquired underground (hydrocarbons, radionuclides, and heavy metals) as well as traces of the original fracking additives and their byproducts. However, within a given disclosure, it is usually reported simply as “water.” It is impossible to know how often wastewater is reported as water, but to get a sense, we use a new and optional water source data entry introduced by FracFocus in late 2023. From the 1,400 disclosures reporting those data as of April 2024, close to 24.4 percent of the water used, or 5.9 billion gallons, was “produced water” (Allison and Underhill 2024).

Finally, there is a group of chemicals within FracFocus that consists of materials labeled by TSCA as “unknown or variable composition, complex reaction products or biological materials” (UVCBs) (Figure 2D). This category is defined by what is not known about it: by its unknown and variable composition. UVCBs form a large group of materials in FracFocus: up to 20 percent of all identified chemicals. At least one UVCB is present in 94 percent of disclosures, and 50 percent of disclosures have 4 or more UVCBs (Allison and Underhill 2024).

One common type of UVCB within fracking operations is petroleum distillates. Distillates are defined by their production process rather than ingredients: they all begin with crude oil, which is then distilled down to separate materials based on chemical properties such as volatility. Crude oil is itself a UVCB because it can contain hundreds of thousands of distinct chemicals, and its distillates are also often approximate fractions because of variation within the source material, specific conditions within the distillation process, and variability across refineries. UVCBs are known as a particularly vexing health and environmental challenge (Lai et al. 2022) as their impacts are poorly understood.

One of these UVCBs, typically named “heavy aromatic solvent naphtha (petroleum)” and shown in Figure 2D, is in the top 5 percent most frequently reported materials in FracFocus disclosures (Allison and Underhill 2024). This petroleum product is used as a solvent and has the CAS number 64742-94-5. But contained under that single ID number are many individual chemical components that themselves have unique CAS numbers. ExxonMobil's (2011) specification sheet for this product claims it has at least 98 percent aromatics (which are often benzene derivatives and are particularly hazardous), but the SDS for the product (ExxonMobil 2023) lists, at most, half of those components (and this is an unusually informative SDS for a UVCB). Beyond rare exceptions, FracFocus disclosures do not say whether or how much of those sub-ingredients are present.

Structural or Systemic Holes

Finally, there are structural or systemic holes: not invisible, necessarily, but which affect how we can understand the data itself. For instance, FracFocus only includes events that are specifically hydraulic fracturing, but there are many other well stimulation technologies (like acidizing or cyclic steam injection) that use similar but slightly different chemical mixtures. In addition, there's a fundamental disconnect in responsibility for disclosures. The entity required to report chemical use is the fracking operator (the company managing the fracking job), but they often don’t have full lists of chemical ingredients (Horwitt and Teklemichael 2022). Meanwhile, FracFocus, as a third-party entity, claims no responsibility for errors in the data but cannot release operator contacts to the public. The only entity who does know the full chemical “recipe” is the manufacturer—usually mega-companies like Halliburton, which often employ a raft of intellectual property legislation and court precedent to avoid sharing those contents with the public, and even fight against sharing that information with medical personnel and first responders (Wylie 2018). In many cases they will be ignorant of the composition of UVCBs too.

In addition, the “systems approach” is a new reporting format within FracFocus, first introduced in 2016, in which operators can submit chemical constituents separately from the trade names to which they belong and the functions they play. Considered a compromise between public health and fracking companies’ intellectual property, it protects the actual composition of products as intellectual property while reporting individual compounds for emergency response, medical, and environmental health purposes. However, research has demonstrated that without an explicit ban on proprietary chemicals, the systems approach as a voluntary option does not actually reduce proprietary claims (Trickey, Hadjimichael, and Sanghavi 2020). The systems approach also inhibits research that requires knowledge of the link between individual chemicals and the products they’re used within, such as tracking which suppliers use proprietary claims or hazardous chemicals most frequently.

Finally, one might assume that once we have an authoritative chemical identification for a material, we could use that to learn of its potential environmental and health impacts. We have already demonstrated that this assumption does not hold for some specific chemical IDs (i.e., water and all UVCBs) because those imply other chemicals that are not—and likely cannot be—disclosed. Another dimension of this problem is the ongoing lack of knowledge about the environmental and health impacts of even clearly identified chemicals. Some of the chemicals used in fracking can be found on existing lists of chemicals of concern and/or have information about their hazards listed in the EPA's (2023) Cheminformatics database. This is an example of available information that is not included in FracFocus but has been added to Open-FF. Many more chemicals, however, have “no data” in their Cheminformatics profiles, meaning not that they are safe but that there is not adequate information about their safety across multiple health endpoints. This reflects how slow toxicology research processes are in relation to the speed of chemical experimentation, manufacture, and use.

What Do the Holes Tell Us?

Trade secrets are the most visible kind of absence, but as our analysis has shown, FracFocus data is more characterized by a range of other unknowns whose very presence is unknown but whose effects propagate in widening circles through groundwater, soil, air, and human and nonhuman communities. In particular, our analysis highlights the propagation of these impacts far beyond what one set of regulations is ostensibly “about.” Disclosure based on SDS sheets, which only report hazardous ingredients related to occupational health and safety, not only precludes knowledge about which chemicals are used (including whether environmentally hazardous chemicals are used), but also precludes knowledge about chemical transport and effects on human and ecological health. Another example is the interplay between trade secrets and the systems approach response. Trade secret chemical use has historically been understood in terms of intellectual property by US courts, which have tended to weigh their economic value much more heavily than the public right to know (Fink 2018). In this regard, they have been part of a broader corporate strategy using information law far beyond its intended uses: ostensibly protecting investments in innovation, it also puts up roadblocks to lawsuits by preventing any information about chemical use or health effects from entering the public record.

What, then, would a more responsive and useful disclosure system look like? Transparency, in and beyond our US context, only works when users have access and literacy about the information, and when disclosers might actually respond to accusations of poor performance (Mol 2010). Better disclosure design would be informed by its specific audiences and their goals, what end users might require to impactfully use the data, and how to effectively communicate the data in response to users’ accessibility requirements (Konschnik 2014). More concretely, a responsive disclosure system could connect data to maps of the fracking sites, link data to databases of chemical health effects, allow users to calculate the cumulative chemicals used in a given area, or calculate the total mass of chemicals used in a frack or set of fracks. It would function with independent quality control and oversight, be reviewed and used regularly to identify and address problems (at both the individual disclosure level and at the level of the system), and would be publicly available and easy to use. It would allow permitting agencies to account for all the other fracks in an area before deciding on the next permit, have response mechanisms that trigger complementary regulatory requirements, and it would serve as a platform from which affected communities could learn about existing impacts and make meaningful decisions about their lands’ futures. Finally, disclosure frameworks don’t function within a vacuum; a better disclosure system would require more responsive and reciprocal relationships across the landscape of environmental regulation and permitting. Transparency is only as radical as it is made to be.

Even while we call for better disclosure frameworks, our most important finding is that chemical worlds—and embodied imbrications with them—are much more complicated and less contained than chemical boosters and critics might want to acknowledge. Both UVCBs and produced water, for example, are holes that cannot be filled even with good will and better data; these two holes in the fracking disclosure data reflect the uncontrollability, fluidity, and movement of chemical use and behavior. In other words, it isn’t just that disingenuous disclosure regimes provide bad data (though that appears to be part of the story in the United States), nor is it just that these data—like all data—are partial. Rather, the point is that disclosure as the primary regulatory response is premised on the idea that disclosure can be complete, and that complete knowledge is a regulatory end in itself.

The absences in FracFocus demonstrate that no matter how much chemical suppliers are pressured to disclose, or how much they ultimately do, disclosure frameworks will run up against the unknowability, variability, and complexity of compounds (far beyond those explicitly named UVCBs) that in their very nature exceed one-to-one, compound-to-CAS-number naming systems. Similarly, because the composition of “produced water” depends on purposefully added ingredients in previous fracks and also their interactions in the environment, including at enormous pressures at subterranean depth, fully characterizing the composition of produced water requires thorough sampling and testing after extraction. More likely, full characterization is impossible. In other words, there will always be holes in what is knowable about the chemicals involved in hydraulic fracturing, and these holes exist only partly out of intentionally obfuscating systems and only partly out of what is unintentionally overlooked or structurally produced as invisible. Chemical behavior itself is definitionally unruly and—in real world settings—unpredictable.

If the holes cannot be filled, the knowledge cannot be recuperated, then what is the power of the desire to do so? From an industry standpoint, there is a profitability to seeing chemicals as contained and containable: even appearing as a “bad actor” that withholds information might instill less distrust than admitting the foundational lack of control that industries have over their products and over chemical behavior. This also puts the often-exclusive focus on trade secret chemicals in a different light: attending mainly to these very visible and quantifiable holes narrows the question of the unknown to one of industry withholding while ignoring the much larger challenge of chemicals’ mutability and mobility. For larger publics, too, the idea that chemicals and their behavior cannot be fully known can be an unwelcome insight that runs against the grain of decades of industrial and chemical engineering. Some environmental justice activism and scholarship has also inherited terms and discursive frames within which there is a possibility for complete knowledge and effective chemical control.

Therefore, these holes ask for a different kind of relearning: about the interconnections between humans and more-than-humans, environments and chemical atmospheres, soil and water, water and air, bodies and their surroundings, and the impossibility of full knowledge and control (Boudia et al. 2021; Cram 2023; Liboiron 2021; Murphy 2017; Nash 2007; Tuck 2009). This necessity travels far beyond fracking chemicals themselves, across the realms of pesticide exposure, nuclear radiation, and the production of “waste” as an idea and a material phenomenon. As chemicals come into being through various petrochemical supply chains, move through fractures deep in the ground, and are “managed” through recycling or disposal pathways, they demonstrate the need to unsettle an assumption of discrete chemicals, discrete bodies, and discrete environments: to go beyond a desire for more data or more complete knowledge toward a fuller understanding of the inextricability of their connections.

CODA: CNX is Earmarked for Environmental Justice Funding?

As CNX's campaign in Pennsylvania unfolds, it remains an example of the embrace of disclosure in the United States, with its expectation that transparency will “definitively confirm” the absence of human health risks related to natural gas production. In a recent annual Sustainability Report, CNX (2024, 32) wrote, “Community members and concerned residents expect transparency and fact-based data regarding the impact of our operations;” and, therefore, “CNX has responded with a historic commitment to further expand our operational disclosures.”

A year after they began their “radical transparency” campaign, and as we were finalizing this article, CNX published their initial results, which they claim showed that its fracking operations were “safe,” “posed no public health risks,” and “inherently good for the communities where we operate.” Yet, as others noted at the time, CNX chose only 14 wells to monitor (out of more than 500 gas wells across their US operations) and measured only a small number of potential air pollutants (Bense 2024): holes in the disclosure indeed!

Moreover, because of their “radical transparency” campaign, CNX was slated to receive federal environmental justice funding through the Justice40 program (Marusic 2024), which aimed to direct 40 percent of the benefits of US federal investments in energy futures to environmental justice communities within the country (Smith 2024). This is the same company that has received over 400 air quality violation noticessince 2020, and pleaded no contest to criminal charges of misreporting air pollution in 2021 (Frazier 2021; Smith 2023). CNX is an excellent example of how transparency and disclosure are still weaponized, and how the simple performance of transparency becomes a tool to demonstrate good stewardship—to the point of receiving federal funds earmarked for environmental justice. That this “radical transparency” campaign is rewarded for providing disingenuous disclosure serves as more evidence that disclosure is being used as a performance of responsibility that can yield substantial material and reputational returns for polluters.