Deconstructing the crash

By Felix Salmon
May 7, 2010
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:

" data-share-img="" data-share="twitter,facebook,linkedin,reddit,google" data-share-count="true">

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.


We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see

thanks for the shout out, felix.

but why would we try to solve this with a tax?

why not a much simpler solution: we have trading curbs for index arb – why not have them for single stocks too? there’s no reason why we couldn’t have a 30 second, 60 second, or 5 minute PAUSE in trading in any stock that declines more than, say 25%…

Posted by KidDynamite | Report as abusive

A tax might be a good idea but I doubt it would ever fly. The industry would fight it with everything they got and I’m sure retail investors would get enlisted to join with them, even if it would be in the retail investors’ interests to have fewer intermediaries siphoning off a slice.

Posted by wpw | Report as abusive

Felix you are on the money in terms of a tiny tax on transactions more than you will ever know. Keep beating that drum. That last paragraph should be required reading for every member of Congress.

High frequency trading is a scam which does not add liquidity but rather serves to allow algorithms to figure out the other side’s limit price and skim the difference.

“instead of providing liquidity, they intentionally probed the market with tiny orders that were immediately canceled in a scheme to gain an illegal view into the other side’s willingness to pay” — Denninger

This rigged market may cause large investors to seek other exchanges besides the NYSE. The perception (and reality) of a rigged marketplace will drive investors away.

The unfortunate upshot of yesterday is that more investors will use limit orders and be subject to the nickel and diming fraud of this high frequency trading.

Posted by DanHess | Report as abusive

Hi Felix,

I agree with KidDynamite. Both Tobin taxes and uptick rules would be inefficient methods to wring out this sort of error. This sort of error could just as well occur in the presence of a Tobin tax with the only difference being that during normal market conditions, bid-ask spreads and other transaction costs would be higher.

One place to look might be the Tokyo Stock Exchange (TSE). The TSE has trading curbs for individual stocks. There are both overnight price change limits and intraday price change limits that place stocks into what they call a “special quote,” (basically an auction period) which can last as long as it takes to stabilize the market.

The TSE rules can be found here in the FAQ (in English): ockprice/p_k.html

Lastly, you write that these algorithms broke during a tail event and that they will be bailed out by rescinding the orders. I believe most HFT firms made out well yesterday. Most HFT strategies typically fare better during periods of higher volatility, and I’d venture that rescinding these trades will have little effect on most HFT firms.

Posted by agathocles | Report as abusive

Don’t we already have a tiny tax on transactions? My broker adds a few cents for “sales fee,” I think they used to call it “SEC fee.”

Not that I oppose a transaction tax, but how could it be enforced? Wouldn’t some shysters come up with a way to agree to pretend that they’d traded, and settle up at the end of the month, reporting only the net?

Posted by rentpayer | Report as abusive

Used to be a NYSE specialist. I don’t know much, but I do know this…

The NYSE was specifically NOT the problem. As it relates to market structure, the problem is that LRPs are unique to the NYSE trading value. When things get wacky, the NYSE floor suspends algorithmic executions, reverting to a simple, yet sensible, human to match buyers and sellers. The system might slow down, but you’ll never see the type foolishness we all witnessed yesterday.

The LRP is valuable for listed companies and many market participants, but absurd if only a feature of one trading venue. Kind of like financial reform in one country. The SEC should mandate that all trading venues that report (in real-time) to the consolidated tape have LRP-like trading curbs.

Posted by kriscollins | Report as abusive

Currently, with no transaction tax, HFT firms can earn a profit making a market on a stock with a 1 cent bid-ask spread. If the government put in place a 1 cent transaction tax, why wouldn’t the bid-ask spread just increase to 2 cents, with the HFT guys still making their 1 cent? When does the bid-ask spread get wide enough that human market-makers have an edge over HFT bots? Does it ever get wide enough? Or would the bots always beat the human market-makers to the punch and take their 1 cent, and the rest of us (investors) would pay the transaction tax in the form of wider bid-ask spreads?

Posted by o_nate | Report as abusive

Felix, any further investigation into the true source and deconstruction of this would be appreciated. As I remarked to a colleague, it’s interesting how the most important financial reform bill in ages is going through Congress right now, and elements of Wall St are consistently shooting themselves in the foot.

Posted by REDruin | Report as abusive


You decry “bad liquidity” as the sort that “dries up to zero just when you need it most.”

But why would any sensible market-making human (or algo) offer the kind of liquidity you want — “just when you need it most” — at the same price (same bid-offer spread) as when it is NOT needed most, when it is NOT a “fast” or momentary crisis-likemarket?

It sounds like your criteria for distinguishing “good” from “bad” liquidity is whether or not one can get a tight, deep, two-sided market at ALL times, under ALL circumstances.

No market-maker, human or algorithmic, will do that, ever.

Posted by dedalus | Report as abusive

En passant, liquidity is nice when you have enough of it to stay in the game, no matter who you are. Goes without saying, I know. Too bad too much of it has been placed out of reach for too many by so few. Yesterday may have been merely a warning shot in the financial corporate battle over the final fragments of humanity’s disposable income.

One of the numerous odious features of the (hopefully, temporary) post Glass-Steagall era is that those having synthesized the greatest liquidity are making it their business to steadily reduce everybody else’s to zero, crash by crash… because they can. That’s not a market worth saving, no matter how plaintive the TBTF supplication after they’ve wasted everybody else’s wealth.
Strictly speaking, it isn’t a market at all, more like a gated community full of parasites.

What happens now in our HFS Terminator economy is like killer robots on eBay determining the price of everything, but knowing the value of nothing. Life on Wall Street has become a plagiaristic heuristic imitation of that art piece of which you recently blogged. As an art piece it’s entertaining and educational – as an actual way of life, psychopathic.

While the actual line where Kid Dynamite’s reasoning ends and the real craziness began may be too blurry to determine, it’s evidently already been crossed, the market shamelessly out of control. What happened yesterday, more like an economically-transmitted Wall Street virus than a bug, is likely to replicate without further warning.

As its equal and opposite reaction, what just about everyone not indentured to one or other of the radioactive Darth Vader death stars of Wall Street now wants most of all is to see the whole shooting match not just regulated, but totally delenda-est-ed.

Posted by HBC | Report as abusive

Interesting…. however, lets suspend disbelief for a second to consider the following. Could the stock market have been hacked? I find the unusual pricing of certain “stable” stocks to be highly irregular. I dont think the HFT and trading algorithms would have led to such a decline. Also, why those stocks, why not other ones, or all of them? We should also be looking at the stocks that actually ended the day higher…why were they immune?

Posted by LucidOne | Report as abusive

Can I place an order that says, “execute this only on NYSE/Arca”, rather than seeking out NBBO? If not, does it even make sense for different platforms to seek out sensible trading rules? If NYSE implements a trading pause — which seems like a good idea — and that causes my order to get routed to an ECN that doesn’t, that isn’t good for me, for NYSE, or for the stability of the market in general. There might well be some people who would always say, “NBBO”, but we need macroinstitutional rules that allow for competition among exchanges to create good microinstitutional rules. An NBBO mandate may well short-circuit that.

Posted by dWj | Report as abusive

A fat-finger and some poorly coded algos could certainly trigger a meltdown.

Suppose you have 6 big algorithmic orders to sell significant positions in a stock and try to participate in 20% of every trade (VWAP orders – note that it’s hard for 6 people to each sell 20%, whoever is quickest would participate and the one left out would be looking to catch up)

Then suppose a technical level gets breached and a couple of algos say ‘sell now at any price’

If there’s insufficient buying volume and poorly written algos, you end up with a race to the bottom, and you could get stock offered at a penny or less before someone steps up to buy.

So the fat finger plus itchy-trigger-finger algos could certainly wreak some insanity in a nervous market.

Posted by TwasBrillig | Report as abusive

Incidentally, if I look at the S&P 500 on Thursday with Bollinger bands, the 2:30-2:45 period doesn’t look especially different qualitatively from earlier in the day, where the market was bouncing off new lows that were descending in an accelerating way. The liquidity collapse from market fragmentation presumably exacerbated what was going on; beyond that, it looks like a bunch of trades based on technical analysis could well have driven the market to the point where that amplifier kicked in.

Posted by dWj | Report as abusive