Fair Lines Am. Found. Inc. v. U.S. Dep’t of Com., No. 21-1361, 2022 WL 3042188 (D.D.C. Aug. 2, 2022) (Berman Jackson, J.)

Fair Lines Am. Found. Inc. v. U.S. Dep’t of Com., No. 21-1361, 2022 WL 3042188 (D.D.C. Aug. 2, 2022) (Berman Jackson, J.)

Re:  Request for “‘documents identifying the total population (number of individuals) imputed statewide by the Census Bureau for group quarters for each U.S. state’”

Disposition:  Granting defendants’ motion for summary judgment; denying plaintiff’s motion for summary judgment

• Exemption 3:  The court “finds that the agency has carried its burden and is permitted to invoke FOIA Exemption 3 in declining to release the [Group Quarters Count Imputation (“GQCI”)] numbers.”  The court first notes that “[t]he Supreme Court has already made clear that the portions of the Census Act defendants cite are an appropriate basis for the denial of a FOIA request under Exemption 3.”  “So the only question here is whether the withheld material falls within those confidentiality provisions.”  “13 U.S.C. § 8(b) states that ‘the Secretary may furnish copies of tabulations and other statistical materials which do not disclose the information reported by, or on behalf of, any particular respondent.’”  “13 U.S.C. § 9(a)(2) prohibits the agency from making ‘any publication whereby the data furnished by any particular establishment or individual under this title can be identified.’”  The court relates that “Plaintiff points out that its FOIA request, as narrowed, does not ask for the ‘information reported’ or ‘data furnished’ by any particular individual or respondent.”  “Instead, plaintiff is requesting aggregated data:  ‘the total population (number of individuals) imputed statewide by the Census Bureau for group quarters for each U.S. state.’”  “And defendants acknowledge that this request, on its face, would not require them to disclose the information or data reported by any particular individual or respondent.”  “The issue, according to defendants, is the impact that this disclosure would have on the strength of the larger dataset.”  “They present a lengthy description of their plan to protect ‘the underlying raw data’ from being ‘reconstruct[ed]’ ‘through a series of mathematical algorithms.’”  “Central to this process of protecting the census dataset is the allocation of ‘precise amounts of statistical noise’ while maintaining some ‘invariants.’”  “Noise infusion ‘introduces controlled amounts of error or “noise” into the data.’”  “Meanwhile, ‘[i]nvariants are data held constant.’”  “In other words, when releasing information related to the 2020 Census, the Census Bureau publicizes some numbers as they were actually reported to the agency.”  “Meanwhile, the agency publicizes other numbers slightly differently than how they were actually reported – these numbers have been infused with noise.”  “The goal of inserting deliberate amounts of noise is ‘preserving the overall statistical validity of the resulting data while introducing enough uncertainty into the data that attackers would not have any reasonable degree of certainty that they had isolated data for any particular respondent.’”  “Defendants argue that this process ‘protects privacy while maintaining the overall statistical validity of the data.’”  The court relates that defendants claim that “the more data that is publicized as invariant and without the introduction of noise to protect anonymity, the easier it would be for someone to ‘reverse-engineer releases of aggregated data to identify individual data,’ especially in this modern era with the extreme amount of ‘computing power that exists today.’”  The court finds that “the agency’s privacy practices provide context for the legal question, but the direct question here is:  does the Census Act prohibit disclosure of the requested information.”
  
  The court finds that “[t]he first provision, 13 U.S.C. § 8(b), clearly does not prohibit this kind of disclosure.”  “Plaintiff has not asked for ‘raw census data,’ . . . submitted by any individual; the narrowed request calls for an aggregated count of ‘imputed’ individuals – just the total number without any further elaboration or identification.”  “The request is therefore outside of the 13 U.S.C. § 8(b) in two different ways.”  “First, the furnished information would not include information that has been ‘reported.’”  “The GQCI numbers are not reported by anyone; they are imputed, and therefore the source of the information is the agency itself.”  “Second, the request does not ask defendants to ‘disclose’ information that stems from ‘any particular respondent.’”  “Again, the nature of imputation is that the numbers are not tethered to, let alone submitted by, any particular respondent.”  “Because the question under 8(b) is whether the furnished information would itself disclose reported information – and it clearly would not – it is irrelevant whether disclosure could be part of the longer chain of events that defendants describe, in which such information could eventually be discovered by an unrelated third party.”
  
  “13 U.S.C. § 9(b) is written differently, though.”  “The text of § 9(b) prohibits ‘any publication whereby the data furnished by any particular establishment or individual under this title can be identified.’”  “Applied to this situation, the adverb ‘whereby’ means that publication is prohibited if the publication could help ‘the data furnished by any particular establishment or individual . . . be identified,’ or through publication, ‘the data furnished by any particular establishment or individual . . . can be identified.’”  “The situation defendants describe – in which release would contribute to the ability of a third party to reconstruct the dataset – falls within the category of a situation ‘whereby,’ or ‘through which,’ ‘data furnished by any particular establishment or individual’ could be identified.”  “It would be through the publication of additional invariant data that the third party could more easily reconstruct data furnished by a particular individual.”  “This conclusion brings [the court] to plaintiff’s other argument:  that regardless of how the statute is interpreted, the hypothetical attack imagined by defendants is not just far-fetched; it may even be impossible.”  “Plaintiff asks the Court to accept its expert’s opinion on the likelihood of harm the withholding is designed to prevent.”  “It insists that defendants’ efforts regarding invariants and statistical noise are unnecessary.”  “But the question when examining the sufficiency of an agency’s reasoning in a FOIA affidavit is not whether the plaintiff has an equally or even more compelling theory, but rather, whether the agency’s justification appears logical or plausible.”  “The agency has explained its thought process in an extremely detailed affidavit that evinces a good faith belief that the risk is real.”  “This belief is supported by historical examples in addition to theoretical reasoning.”  “Given the agency’s statutory responsibility to protect its own dataset, its greater level of familiarity with all of the numbers involved (including those that have been infused with noise for publication), and its lengthy deliberation about these very risks over the last ten years – including reconstruction simulations and ‘regular outreach with dozens of stakeholder groups,’ . . . – the agency’s good faith determination that the risk is real suffices.”