Flash Crash Explanation
Last weeks 1000 point dive in the Dow was pretty scary and people are still trying to figure out what happened. Here is my best guess.
First a quick review on how markets work. In the good old days of the stock “specialist”, he (and it was almost always a he) would be responsible for making a market for a stock. This meant he would take the limit buy and sell orders and record the various bid and ask prices. When a market order came in, he had a choice, either he could match the order with the complementary bid or ask or he could fill the trade himself. Why would he fill the trade himself?
Imagine a stock has a bid price of $10 and an ask price of $10.05. If the specialist believes that the stock price is not likely to move very much that day, when a sell order comes in he may buy the stock at $10 himself and then when a buy market order comes in he may sell it at $10.05. As long as prices don’t move dramatically between when he buys and sells this makes him a nice profit and it improves the liquidity of the market. The specialist can do this because he sees the trades coming in before they are executed and can then judge the risk of playing the market maker role. If it becomes clear that the number of buys and sells are heavily asymmetrical, he will stop putting his own money at risk.
Now let’s fast forward to today where the world gets a little more complicated.
First, we now have flash trading. Flash trading is where an investment house (e.g., Goldman) puts a computer on the exchange floor and is given access to the market order stream before they are executed (much like the specialist used to get). The programs then have on the order of 1/2 a second or less to decide whether to jump in ahead of any existing bid or ask orders on the books. About half of all the trading volume occurs with these flash programs on one side of the trade. In many ways these programs make their money like the specialists of old.
The other big change that has occurred is that a number of alternate electronic exchanges have sprung up over the years because various large players were unhappy with NYSE rules. So instead of all trades going through the NYSE, there are a handful of other venues you can shop around for the best price. So when you place a trade with your broker dealer, their best execution algorithm is becoming increasingly important. This is the logic they use to decide which exchange to route your order to.
So now let’s go back to Thursday May 6th. People were panicking and the market was moving steadily downward. Two things then happened which hit the accelerator on this trend. First, many of these flash trading programs were turned off. Remember, these programs can only make good money when the price of the stock stays relatively stable and they can play the spread. When it is moving too much in one direction the spread isn’t big enough to cover their losses. And when those programs turn off it means twice as many sell orders are hitting the bids on the books.
The second thing that happened is that because the market was getting panicky NYSE went into slow mode where they process fewer orders per second. The purpose of slow mode is to give humans a chance to process what is going on and react to it. Now here is where the other exchanges and order execution algorithms come into play. A sell market order is submitted. The best execution algorithm asks each of the electronic exchanges for their best bid. Since the NYSE is in slow mode, the other exchanges come in with answers and ability to execute much faster. The algorithm then has to make a choice. Does it wait on NYSE before making its decision or does it plow on ahead with incomplete information. Well some people have made their algorithms fairly time sensitive so they would end up ignoring NYSE and executing on one of the electronic exchanges.
In normal times this wouldn’t be a problem because the flash programs would keep the the prices pretty well synchronized. But with them turned off the trades were going against the very limited bids these alternate electronic exchanges had on their books. Remember the majority of bids and asks are on the NYSE, so it didn’t take a lot of volume before the higher priced bids were exhausted and the much lower priced bids were executed.
Once these low priced trades were recorded, this then triggered many stop orders. The way a stop sell order works is it says, if you see a trade that sells below price $x then submit a market sell order. So even though the original low ball order may have been an outlier on a different exchange. That low trade then triggered many stop orders to convert into market sell orders which just through fuel into the fire and caused things to crash even more.
Once people realized prices were going silly low the bid volume came back and prices quickly snapped back to slightly below where they had been before the crazy part of the crash.
The NYSE then decided to invalidate all trades that were more than 60% below their precrash level. This caused its own problems. If your stock precrash was at $20 and it closed at $20 and you stopped out at $10 you are out of luck (since that was only 50% lower). Furthermore, if you bought that same stock at $6 and sold it at $12, you would find that your $6 trade was invalidated but your $12 trade was not, so you would find your self short and having to cover. So instead of making $6 a share you would lose $8 a share.
So what does it mean? For people who invest in low turnover funds and don’t use stops, problems like this wont affect you much. But for more active traders and those who use stops in hedging strategies, the world has gotten much more dangerous.