Nassim Nicholas Taleb made famous the concept of "tight coupling" to explain why there were sudden, interlocking failures in different markets or sudden crashes in prices of securities.
A recent example highlights and illustrates this problem beautifully. An obscure book on genetics of a fly, The Making of a Fly, created a record on Amazon.com when the price quoted for a used copy of this out-of-print book went up to beyond $23 Mn (shipping $3 extra) as recently as in April this year! (At the time of writing, the price was down to $65)
Apparently, two booksellers who listed this book as among their offerings, had an algorithm (i.e., a computer program) that quoted the price of the books, esp. used books, they offered for sale, so that the price that was quoted was never very far from the market price. Quite independently, the algorithm both used took, among other things, the price quoted by the other as a benchmark, and raised it by about 10%. So, effectively these algorithms competed with each other to set a higher price! It is easy, knowing this, to understand how this caused the price to spiral beyond reason. With no human being checking the price quoted, the price soon went beyond the dictates of reason, with no stopper!
Exactly the same thing happens in stock markets as well. This is especially true in the so-called High Frequency Trading (HFT) firms. Today, well over 70% of the trades by volume as well as value in the US are carried out by HFT firms' computers that initiate orders based on information received electronically, before human traders can even read and process the information they observe, leave alone decide and implement the decision.
Algorithmic trading, or algo trading, or simply black box trading, is also used to divide large trades into several smaller trades in order to manage market impact, and risk. Sell side traders, such as market makers and some hedge funds, provide liquidity to the market, generating and executing orders automatically. Algo trading is also used almost every investment strategy for market making, arbitrage, or pure speculation (including trend following). That is the "good side" of algo-trading.
It is very common for traders to also put in "stop-loss" limits in algorithms - the point to which, if the price falls, a sell order at market is implemented. This is intended to cap the downside of any bet taken. However, there is a major side-effect of this: When prices are falling, when stop-losses are triggered, the number of shares being sold sharply rises - which results in the prices falling further, which then sets off a fresh wave of stop-loss orders ... and so on, till you have a price crash that nobody can stop, because it all happens faster than the human mind can comprehend and act on! This is analogous to how we get pile-ups on high-speed superhighways, but much lesser scale of accidents on very crowded roads.
That is the reason why, today, sudden single-day (or even single-hour) falls in several markets all over the world are unsettling, but alas, not infrequent occurrences.
In the next few days, I will be writing more on algorithmic trading. Look out for more!