Home > Research and expertise > Research highlights > Economy > The rise of the machines?

The rise of the machines?

In automating the financial markets it’s important to retain the ‘social’ component that people bring according to Daniel Beunza.


On 6 May 2010 a Kansas mutual fund programmed a computer trading algorithm to aggressively sell 75 000 future contracts, worth around $4.1 billion, in the space of 20 minutes. This large sale, which would normally be done over several hours, confused other automated, high-frequency, traders in the market. These began rapidly buying and selling the futures back and forth between them, sending prices plummeting.

When prices bottomed, the Dow Jones was down 998.5 points – the biggest one-day point decline in the index’s history. And, because of the speed of the technology involved, it had dropped 600 of those points in just five minutes.

“The ‘Flash Crash’ was a wake-up call about the potential dangers of algorithms," says Dr Daniel Beunza an economic sociologist who is interested in the social dimension of financial markets. "Just a few years before, there had been a consensus that automation of stock markets was the way to go. But the crash called into question the paradigm of the market –  that it was just an information processing device that can be automated.”

Beunza's research, with his colleague Yuval Millo, on the New York Stock Exchange (NYSE) before and after automation challenges this idea of the market as merely a database which matches buyers and sellers. Instead they found that there was a strong social component to trading, even after the Exchange had embraced algorithms.

They focussed on two key roles at the Exchange – the specialists and the floor brokers. Specialists are the main facilitators of trade who match buyers with sellers and make sure there is a continuous market. Floor brokers represent a brokerage house and buy and sell orders for their clients.

In the traditional auction specialists stood on a fixed physical post while floor brokers moved between them to complete transactions in different stocks. Beunza describes following a floor broker from one post to the next: “As we followed … he stumbled upon fellow brokers and specialists whom he addressed, backslapped and saluted with nicknames. Everyone on the floor was Johnny, Jimmy or Bobby – there were no Johns, James or Roberts.”

The ‘social’ extended to trading too. When a broker had a buying order, for example, he didn’t just hand it over to the specialist for processing. Instead a conversation would take place between the two with the specialist revealing selective information to encourage the order but without compromising the positions of existing orders on the books. By the specialist disclosing, for example “Merrill has an interest” or “Morgan’s a seller”, the floor brokers could work out the broad identity of buyers and sellers in the market and their strategies and act accordingly.

Specialists also managed the rhythm of the market and, in so doing, helped preserve liquidity. By managing the crowd of brokers at their posts specialists could temporarily slow down trading and ensure that traders found buyers or sellers at a similar price and thereby minimize the price volatility.

bull_marketAfter an initially unsuccessful attempt to automate in 2005 under the NYSE’s then Chief Executive John Thain, his successor Duncan Niederauer hired a team of executives familiar with the NYSE to better integrate algorithms with its traditional floor auction.

The Next Generation Trading Floor was launched in 2008. It included a re-conceived specialist who was allowed to use algorithms for the first time and was more like other traders. However the specialists’ role in controlling the rhythm of trading in a way that would dampen the ups and downs of the market had been maintained.

When stock became very volatile the algorithm was switched back to manual, incorporating automation's advantages of speed and cost with a human's judgement. Beunza describes the system as being, "like a pilot switching off the autopilot in order to land the plane”.

As part of the new trading floor, brokers were given handheld computers which they use to buy and sell. It was initially used by the brokers to just annotate buying and selling – like an electronic version of the notebooks they had traditionally used. However it has now been developed to include algorithms which a broker can programme to buy or sell shares when the conditions are right, but which can also be stopped or adjusted.

The handhelds also provide brokers with information, previously supplied by specialists, about what else is happening in the market. On one screen allows them to see a list of the badges of other floor brokers who are buyers or sellers in a given stock. Tapping the badge opens up a social networking function that allows brokers to exchange information that would have previously been done verbally.

Whereas floor brokers used to walk so much between specialists’ posts they wore special shoes, they can now stay in their booths and through the handheld, be in every one of the crowds.

“The act of buying and selling on the NYSE have been automated but that doesn't mean that people have disappeared. Judgment, the ability to understand what's going on in the market and awareness of what others are doing have stayed,” says Beunza.

"Tellingly, when the Flash Crash happened, the NYSE system flipped into manual, highlighting there is still a very important role for people, and not just machines, in the markets.”

Posted April 2012

Useful links

For full details of Dr Daniel Beunza's research and publications see his profile on the LSE experts directory: Daniel Beunza

Drs Beunza and Millo are authors of Socializing finance, a blog about the sociology of finance.