Monday, July 06, 2009

Ex-Goldman Quant Arrested

Bloomberg reports a computer programmer, Sergey Aleynikov, was arrested July 3 on arriving at the Newark, New Jersey, airport and charged with theft of trade secrets. As someone involved in intellectual property litigation, it brings back nightmares, though I am happy to say I've never been arrested.

From Bloomberg, they note that:
the proprietary code lets the firm do “sophisticated, high-speed and high-volume trades on various stock and commodities markets,” prosecutors said. The trades generate “many millions of dollars” each year, they said.

This highlights the problem with IP law. These statements sound like sufficient probable cause to a judge, but they could be a program that buys stocks based on the simplest of pairs trading algorithms. The problem is that intellectual property law in most states allows an ex-employer a very broad scope to level charges, so if someone wants to accuse you of something very vague based on the flimsiest arguments(eg, 'he planned to do mean-variance optimization as he did at our firm') they have the right. As an instrument of harassment this is much scarier than any anti-terrorist legislation, because no one cares when it is abused. The fact that while they engage in discovery (which can last years) and send and receive interogatories, searching your hard drives and emails, you cannot work for anyone because without a precise definition as to what information, code, algorithms, etc., are at issue, anyone doing business with you is taking on infinite liability. Plus, good IP lawyers cost a lot of money, and as litigation takes several years, a large firm can easily bankrupt a target. With such a prospect, a rich litigator can make the ex-employee an offer he can't refuse.

Firms who engage in such over broad litigation are invariably alpha-less because only those so insecure about how outsiders view their alpha, and how important non-principals are to their alpha, use this tactic gratuitiously. In the end, it does not work, mainly because 1) they demonstrate they believe they have little endogenous alpha, which is usually true and 2) no new alpha comes in, because those with a choice prefer to avoid such people.

In this case, I think Goldman is probably on the right side. This is because Goldman has a habit of letting ex-employees leave and prosper, as they have learned to use their ex-employee success to their advantage. They have no insecurity that people will think Goldman's alpha is from the current crop leaving to start new companies, as this has been going on for decades. Further, the person at issue is a computer programmer, not a trader or portfolio manager, and such code is often very specific (not broad concepts like 'variance' or 'cash flow'). Lastly, programmers are usually given algorithms to code; it is implausible that such workers already knew about parochial trading rules based on previous work automating account statements for Blue Cross.

1 comment:

Michael F. Martin said...

Somewhat paradoxically, the best way to avoid trade secret disputes (which I agree are messy and often unfair to one party) is to file lots of patent applications. This keeps the parties honest later, and also forces the parties to work out ownership issues upfront, before any $ is won or lost.