Bailout

Deconstructing the crash

By Felix Salmon

Bloomberg’s Nina Mehta and Chris Nagi have an excellent explanation of the role of fragmented exchanges in yesterday’s market crash. The upshot is that something which was meant to make trading safer in fact made it more dangerous, just like portfolio insurance in 1987. And the background is the way in which the big two exchanges just aren’t as big as they used to be, at least on a relative basis:

Increasing automation and competition have reduced the NYSE and Nasdaq?s volume in securities they list from as much as 80 percent in the last decade. Now, less than 30 percent of trading in their companies takes place on their networks as orders are dispersed to as many as 50 competing venues, almost all of them fully electronic.

They thought they could turn this bug into a feature, but it turned out to be a real bug after all:

Rapid-fire orders trigger what the NYSE calls liquidity replenishment points, or LRPs, shifting the market into auctions. While the system is designed to restore order on the Big Board, trading is so fast during times of panic that orders routed past the exchange may swamp other venues and exhaust buy orders, said Angel at Georgetown.

Conceptually it’s a bit of a stretch to hope that in extremis, if the NYSE runs out of liquidity, then the smaller electronic exchanges will be able to provide it. But that’s the idea behind LRPs. With any luck, in the wake of yesterday’s chaos, the whole LRP system will be revisited.

Meanwhile, Kid Dynamite is on fire right now, with three posts getting to the real nub of the Crash of 2:45. He started good, pouring cold water on the “fat finger” hypothesis; that skepticism is holding up, even in the face of an anonymous “official” in the NYT talking about ?a huge, anomalous, unexplained surge in selling”. He then got better, noting where the big correlations were (think the yen carry trade), and blaming not humans with fat fingers but rather algorithms which break during tail events.

And then this morning he explains why it’s a really bad idea to bail out those algos by rescinding the craziest trades from yesterday:

Merkel’s point is simple and accurate: if buyers who step in later see their trades canceled, it removes all incentive for them to step in – and then you don’t get the bounce back that we saw! Think about how much havoc it causes a trader who astutely bought cheap stock, then sold it out at a profit. He’s now short! …

If anyone wants to defend the decision to cancel the trades, I’m all ears – but your argument needs to be better than “HEY ITS NOT FAIR THE COMPUTERS RIPPED ME OFF AND MY STOP ORDER GOT EXECUTED AT A PENNY WTF OMG *$XYS !^^!@&!*#” If you don’t understand that this can happen with a stop order, don’t use stop orders.

There’s a very sensible idea going around that a simple way to deal with nearly all of these problems, at a single stroke, would be to implement a tiny tax on financial transactions. Historically, people have complained that such a tax harms liquidity, which is true. But the fact is that it harms the bad kind of liquidity — the liquidity which dries up to zero just when you need it most. Liquidity, if it’s spread across multiple electronic exchanges and can disappear in a microsecond, does very little actual good, and in fact does harm during tail events like this. Let’s tax it, and raise some money for the public fisc at the same time as slowing down markets and making them think before doing a trade.

Full post as published by Macroprudential on May 07, 2010 (boomark / email).