Blameless Blackrock

January 10, 2014

If you want a good test of whether someone is an ideologue on the subject of bank prosecutions, just have them look at today’s agreement between Eric Schneiderman, the New York attorney general, and Blackrock. If they think it makes perfect sense, and that Schneiderman should have pursued this line of prosecution, and that Blackrock has been behaving badly, then they will never find a bank prosecution they don’t love. Because this thing is an utter farce.

To read the NYT coverage of the deal, the problem was that Blackrock was trying to get advance inside information on analyst upgrades and downgrades:

Analysts’ changing assessments on the public companies they follow can make a stock plummet or soar, so receiving such information ahead of other investors can be highly profitable for traders.

As a result, regulatory rules require brokerage firms to limit the information flow from research departments to prevent the potential for trading ahead of analyst reports.

But if you read the actual settlement document, it rapidly becomes clear that that’s not what Blackrock was doing at all. Although the AG seems to be doing its very best to obscure that fact.

The entity at the heart of this settlement is a quant-shop subsidiary of Blackrock called Scientific Active Equities, or SAE, which manages an impressive $80 billion. SAE, like all quant shops, constantly monitors a large number of information streams, and then trades when the streams display certain pre-set characteristics. It’s basically a set of if-then rules: various patterns trigger various different buy or sell orders.

Most, but not all, of the information streams that SAE monitors are wholly public information: momentum, earnings, trading volume, that kind of thing. But there was one stream which was non-public: a periodic survey that SAE would send out to its brokers. Brokers talk to big investors all the time, of course: that’s their job. And they’d happily talk to any investor working for any $80 billion subsidiary of Blackrock. But SAE is a quant shop, and it’s hard for a computer to have a short conversation with an analyst. So instead, SAE would send out a questionnaire — which was very clear, at the top, that it was asking only for public information, which had already been disclosed in research notes, investor calls, and the like. What’s more, even if that was a CYA disclaimer, which everybody was happy to ignore, the questions in the questionnaire are exactly the kind of questions that a sell-side client would ask of a sell-side analyst on a phone call. And as Matt Levine points out, neither Schneiderman nor anybody else is looking to put an end to such phone calls.

But here’s the thing: if a human was asking such questions, they might be fishing for a hint about a possible future upgrade or downgrade. When a computer asks such questions, it isn’t. The point of the questionnaire was emphatically not to try to get inside information on which brokers might be upgrading or downgrading which firms. Instead, it was trying to get a feel for how analyst sentiment in aggregate might be changing, especially around earnings season. Once you added all of the survey responses up and put them all together, then they received a weight of about 5% in the SAE quantitative trading model. But any individual survey response was negligible. An analyst could have said “I’m going to downgrade Stock X tomorrow”, as an answer to the questionnaire, and the SAE model wouldn’t even notice: that’s not the kind of signal it was looking for. A human would notice and care about such a thing, but SAE’s buying and selling decisions aren’t made by humans.

SAE — and I quote the document here — “aggregated and averaged the survey responses before converting them into signals. These signals were expressed as numeric values that were used in SAE’s quantitative trading models”. (My emphasis.) You can’t get insider information from aggregated and averaged survey responses, it’s impossible. And yet, that notwithstanding, Schneiderman still says that the information from the surveys could have been “used to trade ahead of the market reaction to upcoming analyst reports”. This is insane.

Now it’s true that the survey measured analysts’ sentiment — in aggregate. And because it did so on a regular basis, SAE (or its computers) could tell when sentiment — again, in aggregate — was turning. Sometimes, it takes a while for such sentiment to show up in the form of detailed research notes. And in that sense, SAE could take a position in a stock, expecting that the full change in sentiment wouldn’t be fully priced into the market until after at least a few such notes had been published. Which means that Schneiderman is not really accusing SAE of trading on advance knowledge about specific upgrades or downgrades. He’s just worried, to quote a later part of the document, that SAE “could obtain information not generally available in already published analyst reports”.

But it’s investors’ job to obtain information not generally available in already published analyst reports! That information can come from lots of places, including analysts — who, as Levine says, do rather more than just “spend their days in caves writing lengthy reports that they release once a quarter or so”. Indeed, the act of answering questions from clients — or even just filling out SAE questionnaires — is an integral part of the analysis process: it helps analysts get their thoughts in order, and focus on what the clients think is most important. Analysts spend most of their day on the phone to clients — and all of those clients are receiving information not generally available in already published reports. If SAE also received such information, that proves nothing beyond the fact that it’s a client.

Yet according to Schneiderman, SAE’s surveys “violated provisions of the Martin Act, Article 23-A of the General Business Law, and violated provisions of § 63(12) of the Executive Law”. None of these provisions are quoted, and Blackrock was not asked to admit to any violations. But it seems to me that if the surveys really did violate such provisions, then Schneiderman would have been rather more explicit about exactly which part of which law was being broken.

Now it’s in Blackrock’s interest to have good relations with Schneiderman, and so, at the AG’s request, it has stopped sending out the surveys. (And even before it stopped sending them out, it gave them a weight of zero in its algorithms.) But really, Blackrock and SAE did absolutely nothing wrong. And it’s a minor scandal that Schneiderman is bullying them around and forcing them to cease an entirely legitimate business practice. Efficient markets require investors who put work into gathering information, analyzing it, and acting on it. That’s exactly what SAE was doing. It ought to have been receiving praise from Schneiderman, not brickbats.


Comments are closed.