Errors and omissions in FracFocus

Summary: To earn trust of stakeholders, a disclosure instrument like FracFocus must find and fix data errors and omissions. Unfortunately, FracFocus disclosures often have serious errors that go uncorrected for years. We outline how we find these issues, what measures we take in Open-FF to improve data integrity, and our efforts to alert stakeholders to these problems. FracFocus is a critical data source. Making it reliable will require that the companies and agencies responsible take it beyond mere compliance.

List of Topics

FracFocus is one of the largest sources of data about fracking chemical uses available. Established in 2011, FracFocus has accumulated more than 220,000 fracking jobs published by more than 1,600 companies. Most states with fracking activity now require operators to disclose fracking chemicals directly to FracFocus. The disclosures in FracFocus contain important information about water usage, chemical identity and quantity, product names, and trade secret designations. Many of the chemicals disclosed are considered hazardous or dangerous¹ and are used at often very large quantities². As such, FracFocus should be a necessary part of discussions about fracking patterns, developments, impacts and potential regulation.

A disclosure instrument, like FracFocus, is an institutional structure that reveals information that would be otherwise hidden from many stakeholders³. While there may be many ways to measure the success of a disclosure instrument, one important consideration must be how well trusted it is. A major component of that trust⁴ has to do with errors in the data. Data errors are inevitable due to mistakes in data entry, miscommunication, and other mishaps. Errors and omissions may also creep in due company shortcuts or even intentional misreporting. But when an instrument has obvious data errors that go uncorrected, many stakeholders will be suspicious of all the data.

Obvious errors

Unfortunately, glaring errors are not hard to find in FracFocus:⁵

More errors

The examples above could be dismissed as trivial in such a large database. However, when we look more carefully, errors, ambiguities, and omissions are far too common in FracFocus.

• More than 2,000 disclosures report that sand is the dominant ingredient in the fracking fluid, not water.
• More than 30,000 disclosures do not report total water volume. This prevents us from calculating masses of chemicals in the fracking job.
• More than 5,000 disclosures have the same APINumber and date as another disclosure. It is impossible for users of the data to determine which is the correct one.
• More than 18,000 disclosures have apparent duplicate records, which may lead users to unknowingly inflate quantities of chemicals.
• The reported water volume is inconsistent with the reported mass of that water in almost 2,000 disclosures.
• Over 350,000 chemical records report either zero for quantity or the value is left blank.
• Almost 10,000 disclosures have at least one CASRN that must be corrected to refer to the proper material.

These are often serious errors and omissions that may distort or hide critical information. Users of the data must be aware of these problems and often take action to overcome integrity issues.

How can Open-FF find errors?

While state agencies could check and even audit disclosures by requiring additional information from companies, Open-FF is distant from the data source. Because Open-FF relies solely on publicly accessible FracFocus data, our ability to identify omissions, inconsistencies, and errors is limited. Despite this, we have developed several methods for detecting potential issues.

First, we can look for values that are invalid, such as a percentage that is greater than 100%. We can also look for values that are clearly out of range. One example would be a vertical depth that is far beyond physical possibilities. Another is a published fracking date far in the future.

Second, we can look for system inconsistencies. FracFocus disclosures include some redundant information, so there can be two ways to arrive at one value. For example, the quantity (mass in pounds) of a chemical can often be determined either directly (MassIngredient column) or through calculation using the percentage of the chemical in the whole fracking fluid. Another example is that chemical identity is typically described by both an authoritative number and by a name. When the two sources are in conflict, the underlying data is ambiguous at best, but possibly wrong. Furthermore, we can use independent chemical disclosures published outside of FracFocus to look for inconsistencies. Pennsylvania’s Department of Environmental Protection’s disclosure system is an example of a parallel data source we can use.

Finally, we can examine disclosures in the context of the whole of FracFocus data. For instance, we should be suspicious if one disclosure reports a chemical at a concentration that is million times greater than 10,000 other disclosures. Similarly, we should flag a job reporting water volume of three gallons when a typical value is 16 million gallons. The figure shows another example. The typical percentage of sand in a fracking fluid is about 11%, which highlights the extreme percentages in sand-dominated disclosures.

Feedback requested

Because Open-FF has no direct connection to the companies or their data, it is possible that we have misinterpreted data in the disclosures⁷. When that is the case, we depend on companies, agencies or FracFocus to inform us of the actual situation. Our misreading would imply a misunderstanding of the disclosure instrument – which is purported to be “a one-stop, easy-to-understand public resource for consumers wishing to explore [fracking] data”⁸ – a situation that should be cleared up if the FracFocus is to fulfill its promise.

In addition, we are always interested in hearing from users of the FracFocus data. It is often that data problems are found while carefully inspecting data for analysis. We welcome stories of such problems to help us alert other users.

Making errors visible

Because Open-FF has found these weaknesses in FracFocus and that many are rarely corrected, we have started an effort to make the problems more visible to stakeholders. Our goals for this are to alert users of the data to be aware of the issues and to encourage operators and state agencies to correct existing problems and to prevent future data errors.

When feasible, we contact the operating company of problem disclosures to alert them to the issues. Because FracFocus cannot release the company contact information to us, they have offered to forward our alerts to the contacts directly. In at least one case, after we alerted FracFocus Support to the specific details of one company’s errors, they informed us that they planned to send out an announcement to all affected operators.

Further, data sets generated by Open-FF include accommodations to some of these errors. For example, many disclosures include conflicting or ambiguous chemical IDs. Open-FF is able to resolve many of them into a new column for users. Further, while the mass of a chemical might be available in the FracFocus column, MassIngredient, Open-FF checks it for consistency and compares it to a calculated mass to validate its accuracy. When it detects problems, Open-FF does not report mass. This prevents Open-FF users from unknowingly including erroneous data in an analysis. Users can also easily exclude tens of thousands of duplicated records that were found by Open-FF.

Topics we have posted

In Aug 2024, we began publishing summaries and detailed posts of errors and omissions on this web site. Our intention is to document many reporting issues and to organize these reports at the state level. Topics so far include:

• Sand dominated disclosures
• Duplicated disclosures
• Data omissions
• Water source data

Conclusion

FracFocus is an important public window into an often opaque industry and it is an instrument that the industry supports. For it to fulfill its promise as a key and trustable resource, stakeholders must grapple with its extensive ambiguity, errors and omissions. Those stakeholders using the FracFocus data must stay aware of data traps; Open-FF can help with that. Those stakeholders with the ability to correct errors and change data entry systems – that is, companies and state agencies – should move beyond mere compliance and work to remove problems in past disclosures as well as to prevent them from undermining future data.

---

1. by the Globally Harmonized System (GHS) for Classification and Labeling of Chemicals. ↩︎
2. For example, as of Aug 2024, more than 8,000 disclosures reported the use of more than 10,000 pounds of ethylene glycol, a Safe Drinking Water Act material. About 90 disclosures reported over 100,000 pounds. ↩︎
3. We would include several groups as stakeholders: the multiple companies involved; investors; state and federal regulators, legislators, and commissions; researchers; journalists; community organizers; and citizens both potentially impacted by fracking operations and those just curious. As a public service project, Open-FF tries to facilitate data access for those groups without easy access by other means. ↩︎
4. For a disclosure instrument to be “trustable” by all stakeholders, it must be 1) comprehensive, 2) transparent, 3) auditable and 4) accessible. Because errors are inevitable, correcting errors is intrinsic to the first three on this list. ↩︎
5. All these disclosures were accessed on Aug 26, 2024. If you are reading much later than that, they may have been subsequently corrected. ↩︎
6. in the bulk download column named MassIngredient ↩︎
7. We note, however, that of all the presumed errors we have reported to companies through FracFocus, state agencies and directly, we have received no feedback that we are mistaken. ↩︎
8. Quote from FracFocus.org, accessed Aug. 27, 2024. ↩︎