Financial Cryptography

Goldman Sachs caught with their pants down?

In the world of trading there is a scandal erupting. At first blanche, it was just another insider job. JPMay points to "Goldman: Pwned?". Snippets to get the general thrust. Firstly, GS drops off the trading reports:

This week's NYSE Program Trading report was very odd: not only because program trading hit 48.6% of all NYSE trading, a record high ... but what was shocking was the disappearance of the #1 mainstay of complete trading domination (i.e., Goldman Sachs) from not just the aforementioned #1 spot, but the entire complete list. In other words: Goldman went from 1st to N/A in one week.

Matt Goldstein apparently led the story, reporting that they were inside-hacked, asking "Did someone try to steal Goldman Sachs' secret sauce?"

... a Russian immigrant living in New Jersey was being held on federal charges of stealing secret computer trading codes from [Goldman Sachs].

So, the press made a lot about how this, but it is old news. For the techies reading today, here is how they said they 'got their man':

...the affidavit that Zerohedge has makes clear what they claim they've got this guy cold on - the "bash history" file they're referring to is a Unix system log that the "shell", or command interpreter, automatically keeps. Said alleged offender apparently was aware of this file and tried to erase it after doing his deed, but was unaware that the system he was working on had auditing enabled (oops.)

"Industrial espionage is nothing new of course." Nor is logging the activities of insiders :) This is a standard requirement imposed by audits, whether you like it or not, whether you can do it or not. However, the story grows bigger. This open comment lays it out:

...GS, through access to the system as a result of their special gov't perks, was/is able to read the data on trades before it's committed, and place their own buys or sells accordingly in that brief moment, thus allowing them to essentially steal buttloads of money every day from the rest of the punters world.

The unbacked, unevidenced allegation in the popular blogs is this: the code that was stolen might be been the code that drove a system that "saw" others' trades before they could be executed. More technically, it is claimed:

The big ticket, the magic wand for a rogue quant shop is technology to grab off FIX PROTOCOL, OCX, or SWIFT messages that precede every transaction_commit at the Exchanges.

If true in fact, this is almost guaranteed to lead to front-running. That is, if Goldman Sachs has any such magic wand, they would need to be bona fide arch-angels to avoid the temptation to read the trades coming, and beat them into the market. We're not talking millions here, but billions, or as that blog claimed "The profitability of this split-second information advantage would have been and could have been extraordinary. Observed yielding profits at $100,000,000 a day."

The structural details match: they are given special access for various reasons, and they are on the relevant security committees. Which is perversely backed by the actual claims of the bank:

Assistant U.S. Attorney Joseph Facciponti told a federal magistrate judge at his July 4 bail hearing in New York. The 34-year-old prosecutor also dropped this bombshell: “The bank has raised the possibility that there is a danger that somebody who knew how to use this program could use it to manipulate markets in unfair ways.”

Unlike most other claims written here, this has some degree of reliance. If the prosecutor said it, then, more or less, Goldman Sachs said it. And, the guy from Bloomberg asked the right question:

How could somebody do this? The precise answer isn’t obvious -- we’re talking about a black-box trading system here. And Facciponti didn’t elaborate. You don’t need a Goldman Sachs doomsday machine to manipulate markets, of course. A false rumor expertly planted using an ordinary telephone often will do just fine.

If it is an honest trading system, it can't manipulate the market. Except by volume and sneaky trades and false rumours and so forth, but that is the market. As long as everyone is under the same rules, they are in the market. Goldman Sachs can only manipulate the market if it is in possession of information that others do not have.

The final word seems to go to thefinanser in the UK, which deliciously juxtaposed Goldman's responses to salacious gossip in a popular mag:

The story actually goes back to when Goldmans announced last week that they were fed up with rock mag, Rolling Stone, over a piece claiming that the bank had “engineered every major market manipulation since the Great Depression — and they’re about to do it again.”

Goldmans responded that it was “hysterical in both senses of the word” and that the magazine had “cobbled together every conspiracy theory ever written about us and injected some hyperbole and lots of bad language and called it a story.”

back to back with the prosecutor's filed statements before court:

The problem is that in their case against this guy, Sergey Aleynikov (a Russian immigrant no less, how James Bond is this getting?), the bank's lawyer made the statement that this “raised the possibility that there is a danger that somebody who knew how to use this program could use it to manipulate markets in unfair ways.”

So, the program can be used to manipulate markets, can it?

Goldmans has virtually admitted as much, and no-one is going to let them off the hook.

In fact, GATA (the Gold Anti-Trust Committee) has already raised questions with “the U.S. Securities and Exchange Commission and the U.S. Commodity Futures Trading Commission to investigate the Goldman Sachs Group Inc. computer trading program that, according to a federal prosecutor, the bank acknowledges can be used to manipulate markets.”

What is perhaps worth underlining is that this last post is a credible bunch of insiders, not the rebels-with-a-blog. If the Financial Services Club are calling BS to GS, likely other insiders are too.

And well they should, if Goldman Sachs had an insider track on all the trades into NYSE. If this proves true, we're looking at an event that will make Arthur Andersen look like a Kindergarten squabble.

Audits IV - How many rotten apples will spoil the barrel?

In the previous post on Audits (1, 2, 3) I established that you yourself cannot determine from the outside whether an audit is any good. So how do we deal with this problem?

We can take a statistical approach to the investigation. We can probably agree that some audits are not strong (the financial crisis thesis), and some are definitely part of the problem (Enron, Madoff, Satyam, Stanford) not the solution. This rules out all audits being good.

The easy question: are all audits in the bad category, and we just don't know it, or are some good and some bad? We can rule out all audits being bad, because Refco was caught by a good audit, eventually.

So we are somewhere in-between the extremes. Some good, some bad. The question then further develops into whether the ones that are good are sufficiently valuable to overcome the ones that are bad. That is, one totally fraudulent result can be absorbed in a million good results. Or, if something is audited, even badly or with a percentage chance of bad results, some things should be improved, right?

Statistically, we should still get a greater benefit.

The problem with this view is that we the outside world can't tell which is which, yet the point of the audit is to tell us: which is which. Because of the intent of the audit -- entering into the secrets of the corporate and delivering a judgment over the secrets -- there are no tools for us to distinguish. This is almost deliberate, almost by definition! The point of the audit is for us to distinguish the secretively good from the secretively bad; if we also have to distinguish amongst the audits, we have a problem.

Which is to say, auditing is highly susceptible to the rotten apples problem: a few rotten apples in a barrel quickly makes the whole barrel worthless.

How many is a few? One failed audit is not enough. But 10 might be, or 100, or 1% or 10%, it all depends. So we need to know some sort of threshold, past which, the barrel is worthless. Once we determine that some percentage of audits above the threshold are bad, all of them are dead, because confidence in the system fails and all audits become ignored by those that might have business in relying on them.

The empirical question of what that percentage would be is obviously a subject of some serious research, but I believe we can skip it by this simple check. Compare the threshold to our by now painfully famous financial crisis test. So far, in the financial crisis, all the audits failed to pick up the problem (and please, by all means post in comments any exceptions! Anonymously is fine!).

Whatever the watermark for general failure is, if the financial crisis is any guide, we've probably reached it. We are, I would claim, in the presence of material evidence that the Audit has passed the threshold for public reliance. The barrel is rotten.

But, how did we reach this terrible state of affairs? How could this happen? Let's leave that speculation for another post.

(Afterword: Since the last post on Audit, I resigned my role as Auditor over at CAcert. This moves me from slightly inside the profession to mostly outside. Does this change these views written here? So far, no, but you can be the judge.)

Webmoney's start in the 1998 crisis

In comments, Igor Drokov asked for data points on my claim that Webmoney single-handedly saved the Russian people from their crisis. The problem with Webmoney has always been that the documentation is in Russian, so the story spread slowly and was wildly exaggerated in the telling. I asked Dani Nagy, who is fluent in Russian, for the truth, and here's what he said:

Here is a summary of the official history of WebMoney, as told in 2005 (in Russian) and an interview:

The first financial transaction in WebMoney happened on November 20, 1998, when the shock of the financial meltdown was still raw in Russia. They started their operations with a "Marshall-plan", spending a few tens of thousands of dollars as follows: the first 1000 registered users got 30 WMZ (WM denominated in USD) on their accounts, the first few vendors that signed up for accepting WM got 100 WMZ and invitations were rewarded by 3 WMZ each, if successful.

For about a month, they announced each signed-up vendor as a separate news item on their main page. By December 1998 they switched to batch announcements, as the service was growing in popularity, albeit mostly confined to Moscow due to the (almost negligibly) low residential internet penetration elsewhere in Russia.

The growth was quite rapid. By the end of 1999, businesses operating mostly online, such as ISPs, banner exchanges, hosting providers and web design studios, adopted Webmoney almost universally. It was in 1999 when exchange agents started popping up in major Russian cities. They also got into the remittance business, mostly for Russians working in America's dot com boom.

By 2000, WebMoney was already very popular across Russia. That same year, Oleg Bunas started a branch in Minsk, Belarus. See this (also in Russian).

Of course, in those years, WebMoney was severely constrained by the low Internet penetration in Russia. But among internet users it was a runaway success from the very beginning, as there was no comparable fast and cheap means of payment. The banking sector certainly failed to meet the demand for such.

My (Dany's) comment:

Giving cash to conductors on railroads has been and still remains a popular means of money transfer, but when it's -20C outside (with a raging blizzard to complete the picture), the benefits of being able to wire money from the comfort of one's home or office are difficult to overstate. :-)

The effect of the present financial crisis on WebMoney is thankfully measured by Google.