Trader Talk

What happened during the Aug 24 'flash crash'

Trader on the floor of the New York Stock Exchange.
Brendan McDermid | Reuters

JPMorgan analyst Marko Kolanovic is having his moment of fame.

Kolanovic is the global head of derivative and quantitative strategies. He got considerable attention right after the Aug. 24-25 selloff by publishing a report saying that "price insensitive" programs might cause repeated selloffs.

Yesterday, he got more attention by claiming that "technical selling pressure" from trading firms was largely completed, and that the trend would be toward buying in the days and weeks ahead. The report—out in the middle of the day—was even credited by some with the 200-point rally in the Dow.

Thursday's report also contained a long section analyzing the events of Mon., Aug. 24, when the Dow dropped roughly 1,100 points in the first five minutes of trading.

Kolanovic blamed a good part of that drop on a lack of liquidity in the marketplace, as well as to a herd mentality by traders desperate to buy protection against further market declines at any cost.

Kolanovic notes that the seeds of the selloff were planted the night before. On Sunday night, Aug. 23, a large drop in equities in Asia triggered a drop in index futures in Europe and the U.S. U.S. stock futures went down (7%) prior to the U.S. open.

Panic buying of protection. That, Kolanovic says, caused traders to panic and seek to buy protection in any form, including put options. Lots of it. They got this protection from dealers, who were overwhelmed with the need to sell protection to traders.

When a dealer sells a put option, the dealer is long the market. If the market moves down and the option goes "in the money" the dealer needs to sell to stay neutral. Normally this does not matter, but when you have massive amounts of options hedging that hedging itself will move the market.

These long positions from the dealers got larger and larger, Kolanovic says. Those dealers, to protect themselves, needed to "dynamically hedge", meaning if the market moved down, they needed to sell as well.

And that was a part of the problem: as the market opened down big, dealers were forced to sell as well.

Lack of liquidity. A second problem was a lack of liquidity, which means that there wasn't enough traders willing to buy all the stuff everyone wanted to sell: "a crash would not have happened if there was sufficient liquidity in the marketplace to absorb these flows," Kolanovic said.

How did that happen? A good part of the trading and hedging occurred in the pre-open hours, when there is poor liquidity. Also, the last two weeks of August is typically light on volume.

In other words, traders dumped large amounts of stock at exactly the wrong time.

Why? Why sell into no liquidity? Was there some fever that suddenly struck the trading community that caused them to make colossally stupid sales when there was no liquidity.

Well, in a way, yes. There was a fever. They had to sell, or at least that is the implication.

Look at this way: you are a trader responsible for managing risk at your firm. Suddenly you have outsized risk. You are through your risk limits, and your job is to stay in your risk limits. You are in danger of losing a lot of money, and worse, you are in danger of getting fired if you do nothing.

So you need to flatten your books, which, in plain English, means selling. In this somewhat panicky situation, they are not necessarily looking around considering the health of the market or the participation of other traders.

Market makers and high frequency traders. Kolanovic also briefly mentions the role he believes market makers and high frequency traders (HFTs) may have played in the decline.

He notes that there were extreme dislocations in the market going into the open and in the first 15 minutes or so:

1) Only about half of S&P 500 stocks were opened on NYSE by 9:35 a.m.;

2) 765 stocks in the Russell 3000 were down more than 10 percent on an intraday basis;

3) There were 1,278 trading halts for 471 different ETFs and stocks.

Because of this, it was not possible to calculate the value of many ETFs, or hedge or trade ETFs and stocks at a 'correct' price.

He notes that HFTs and other high-speed traders have models that essentially shut down their systems when they detect extreme pricing anomalies that may be incorrect or erroneous. This is a safety design that essentially says, "Can some human take a look at this and see if this stuff is pricing correctly?"

This certainly happened that morning. Many market participants withheld liquidity when they weren't sure what the prices were, or weren't sure if they were correct.

Is Kolanovic's analysis accurate? I think his key points about a rush to buy protection and a liquidity crunch are indeed components that exacerbated the decline.

However, don't kid yourself—the markets were going down, first and foremost on the fundamental issues of weakness in China and uncertainty about the Fed's policy on raising interest rates.

In a sense, this is oddly reminiscent of the 1987 crash report—which many blamed on "portfolio insurance." Markets were going down no matter what, and we are really debating about collateral effects that relate to liquidity and market structure.

What can be done? That doesn't mean we should ignore the fact that the market frayed around the edges. What can be done to tweak the system to make it work better on days of extreme volatility, like Aug. 24?

Kolanovic does not go into a discussion on this, but based on my own discussions with dozens of market participants in the past few weeks, I have a few observations.

They fall into three buckets:

1) improving liquidity. Kolanovic is certainly correct in noting that lack of liquidity was a big problem.

a) Almost everyone agrees that the volatility would have been lower if all the exchanges had opened at the same time.

One major problem was that parts of the market were open, but other parts—many of the NYSE listed stocks—took several minutes to open.

This allowed NYSE-listed stocks to trade away from the NYSE at widely different prices.

This is an age-old debate. The NYSE still uses a hybrid model, employing electronic trading and floor based designated market makers (DMMs) to open and trade stocks during the day.

So the argument boils down to this: should we just allow an all-electronic open everywhere?

The NYSE has historically opposed this, arguing that investors get fairer pricing in times of high volatility by having humans price the open, and if that means it may take a few more minutes past the open, so be it.

The NYSE has argued that the "just get it open" mentality can be dangerous, that by waiting to get better pricing retail investors—who are the ones who typically get the opening price—would be better served.

Would the markets have had less volatility on August 24th had there been an all-electronic open? I don't know, but I do think it is likely that the system-wide circuit breaker...a decline of 7 percent in the S&P 500....would have been triggered. That would have created a 15-minute pause.

Would that have calmed the market quicker? I don't know. Perhaps.

But it's unlikely that the exchanges will all open all their stocks at the same time; the markets are simply too competitive. No one is going to wait for anyone.

b) how can market makers be prudent, but more active? Kolanovic talked about market makers and HFTs shutting down when the data is can they get more certainty and not shut down?

One thing that's very important to understand about market makers—whether they are on options desks or HFTs or NASDAQ broker/dealers or Designated Market Makers (DMMs) on the NYSE floor—is that they play a game of pennies. They eke out small profits every day.

But history is littered with market makers who went out of business on disastrous days, when they committed massive liquidity and got killed.

So the retiscence is understandable. The obligation of market makers to provide "fair and orderly markets"—an obligation that has been considerably diluted over time—does not include the obligation to go out of business.

c) can the opening rules for the NYSE be improved? Many DMMs had to open stocks manually on Aug. 24. This was no problem when most DMMs had only three or four stocks, but it's difficult to get everything open in a timely fashion when each DMM now has 40 or more stocks. What could help them open stocks quicker, but still fulfill their obligation to find the best price?

2) improving regulation. After the 2010 "flash crash" the entire industry created individual stock circuit breakers, known as "limit up, limit down" (LULD) that halt trading in stocks for 5 minutes when they move more than 5% percent in a rolling five-minute period (the band is widened to 10 percent in the first 15 minutes of trading).

Those circuit breakers have worked well, but on an extreme day a huge number of halts (more than 1,200) definitely caused a problem.

Is five minutes the right amount for a halt? Should it be shorter? Is there a better metric that can be used? For example, why just use a time period if there is the same imbalance of buy and sell orders? Why not use a metric that says, we have to wait until a certain amount of buy orders are available before we reopen when there is an excess of sell orders?

c) changing the practice of market orders and stop orders. Dumping market orders into the exchanges was a problem on Aug. 24. There were many market orders that got executed at prices way below the prior day's market.

People who put in a market sell order on Aug. 24 were understandably upset when they sold down 10 or 20 percent. You can almost hear them say, "I wanted to sell, but I didn't want to sell down that low!"

What could be done? Market orders and stop orders that become market orders arguably should no longer be used. Instead, every order should have a limit, even if it's 10 percent away. That way, no one is surprised.

I'm sure there will be other recommendations. The SEC's Advisory Committee on Market Structure will also undoubtedly address some of these questions in the months ahead.

Whatever is done, let's not lose sight of the fact that August 24th was a very unusual day. There was something akin to a genuine panic among market participants in the pre-open, and nothing was going to prevent a big drop when the markets finally did open.