The NSA’s controversial call records program is predicated on the extraordinary premise that the phone logs of nearly all Americans are “relevant” to an intelligence investigation—and therefore available subject to the §215 business records authority—if algorithmic analysis of the entire database can be used to identify the minuscule subset of those records that actually pertain to the targets of an investigation. Marty Lederman’s recent post does an excellent job of showing how, at a general level, that argument fails.

It may, however, be illuminating to look a bit more closely at some of the specific cases cited in support of this expansive understanding of “relevance,” if only to underscore how radically the NSA’s blanket production orders depart from even the broadest subpoenas authorized by courts in other contexts. Orin Kerr provides an excellent summary of several of the cases cited in the Obama administration’s white paper on §215, concluding that they provide, at best, only very weak and indirect support for the government’s position. That post is essential reading for anyone trying to assess the arguments advanced in that paper.

I want to consider another case discussed at some length by David Kris in his thorough new paper “On the Bulk Collection of Tangible Things,” because Kris suggests that it is the “closest analogue… albeit on a much smaller scale” to the bulk telephony orders issued by the Foreign Intelligence Surveillance Court. In that 1987 case, In re Grand Jury Proceedings: Subpoena Duce Tecum, the Eighth Circuit rejected a motion to quash a subpoena for two years’ worth of Western Union wire transaction records from a particular facility, even though these might include records of wholly innocent transactions.

In one sense, the analogy to the NSA program is obvious: The court understood that while only some of the specific transactions contained in those records would relate to the narcotics investigation for which they were sought, there was no way of knowing which transactions were relevant in advance. It was in order to make that determination that the government sought the whole pool of records.

The more closely we look, however, the more strained that analogy seems. Police in that case had recovered applications from “dope houses” showing several large wires from one specific Western Union facility at the Kansas City Royale Inn, and had additionally learned that several suspects involved in a narcotics trafficking operation had “transferred funds every two or three days to Florida or California” by wire. Prosecutors therefore sought monthly activity summaries from that facility, as well as more detailed records for transactions in excess of $1,000. In rejecting the motion to quash, the court pointed to these features to distinguish this case from several others in which they had rejected excessively broad production orders—noting in one such case that it is “contrary to the first principles of justice to allow a search through all the respondents’ records, relevant or irrelevant, in the hope that something will turn up,” which strikes me as a fairly apt description of the NSA’s bulk acquisition. Here, rather, the court stressed:

The compass of the subpoenas issued to Western Union is nowhere near as sweeping as those in the cases cited above….The type of documents sought is precisely limited to those recording transactions of one thousand dollars or more occurring within a relatively short period of time. Only a single office is involved.

The elephant in the room here—and perhaps one it is just as well the court failed to emphasize, given the unfortunate extent to which effective privacy protections already depend on class—is that online reviews pretty unanimously characterize the Royalle Inn as a seedy, dilapidated motel in a rough part of town. While the monthly summaries requested would no doubt include limited information about many innocent transactions, more detailed records were sought only for wires of $1,000 or more—which would be over $2,000 in current dollars. It does not seem too great a leap to infer that there would not be an enormous number of such wires unconnected with any criminal enterprise at this particular location. Without wanting to endorse any kind of general double standard for subpoenas targeting businesses in low-income neighborhoods, this one does seem to have been deliberately calibrated to limit the collection of irrelevant records, even at the cost of missing smaller illicit transactions, in a way the FISC’s telephony orders are emphatically not.

Finally, as Kris notes, the court ruled only on the question of whether the subpoena was excessively broad under the Fourth Amendment, leaving it to the district court—and, indeed, even encouraging the district court—to further narrow it under the rules of criminal procedure:

The district court may therefore wish to consider the extent to which the government would be able to identify in advance those patterns or characteristics that would raise suspicion. These might include wire transfers to or from individual suspects, transfers to certain locales known to be sources of high volumes of illegal drugs, or other particular patterns designed to focus on illegal activity without taking in an unnecessary amount of irrelevant material.

Such an inquiry is appropriate to protect against unduly encroaching upon the expectations of innocent customers that their financial records will be kept confidential.

If this is indeed “the closest analogue” to the FISC’s §215 telephony orders, it is mostly remarkable for how distant it nevertheless remains.

Stepping back from the details of this case, it is clear that defenders of the NSA program can find instances in which a subpoena or court order compelled production of a collection of records, not all of which were relevant to a particular investigation, in order that the relevant subset could be identified. The trouble is that there is clearly some point at which an after-the-fact relevance determination crosses over into an impermissible fishing expedition—which is why, every now and again, a subpoena does get quashed for overbreadth. And as the NSA might say, you’ve got to look at the entire data set—what the courts forbid, or make clear they would forbid, as well as what they allow—to meaningfully limn the boundaries of “relevance.” Once you do that, it becomes clear that wherever the precise boundary lines might be—and I hope to try sketching some in a future post—you run into them well before you get anywhere near a blanket order for every American’s cell records.