In 2007, the smash-hit game BioShock told the story of Rapture, a city built on the Atlantic seabed dedicated to the pursuit of the free market. Now academic Alexander Wissner-Gross has revealed how the race for light-speed trading could fuel the development of something remarkably similar.
High frequency trading (HFT) exploits tiny differences in the price of a commodity across two markets. As these discrepancies can last a fraction of a second, trading is carried out by computers that make thousands of transactions in milli- or even microseconds. At these speeds, the time it takes to communicate between stock exchanges becomes a major factor in the success of a trader.
The demand for ever-lower latencies have pushed 'tick times' between trading hubs close to their physical limit. Last year, telecoms company Hibernia Atlantic announced plans to lay an optic fibre cable running the shortest possible route between London and New York to achieve a latency of 60 milliseconds. The same journey at the speed of light takes 52.2ms.
In a paper published in Physical Review, Wissner-Gross and co-author Cameron Freer suggest a way to shave that latency even further. "The amount of time it takes light to carry information between geographically separated markets is a limiting factor for trading," he says. "What we derived is the optimal position between any pair of geographically separated exchanges from which to co-ordinate paired trading".
Lying between New York and London, an intermediary node could listen to market prices on both sides of the Atlantic, and initiate trading at either end when prices fluctuated. The banks of computers needed to manage trades could be housed on anything from a floating offshore platform to a high altitude balloon.
The pair calculated the optimum geographic mid-point between 52 major trading exchanges scattered across the globe. Many of these nodes fall on underdeveloped areas in Africa, central Asia, and South America, and this is what Wissner-Gross is most excited about. "Network latency minimisation is already driving new network deployment," he says, "I think that geographic position may become a new form of natural resource that can be exploited by local economies". Just as states levy fees on pipes transporting oil and gas through their territories, taxes could be raised on the flow of information through geographically-optimal locations. But the true value of the nodes is they encourage the dissemination of high-speed trunk lines into areas that would otherwise remain off the global grid.
However, mathematician and economist Leigh Caldwell urged caution on visions of wired cities springing up along technological ley lines: "Depending on the processor speeds of the mid-Atlantic server farms and the time it takes to execute a transaction, they may not be able to issue the buy/sell instructions quick enough to make a difference to the speed of news transmission," he said. "So overall it's an interesting theoretical paper but, I suspect, has limited applicability."
- Log in to post comments
Wouldn't work for geosynchronous satellites since they are 33,000 miles up but for short periods (seconds) satellites in low earth orbit might work and when we have a spacegoing civilisation we could have thousands of them.
Seasteding appears quite possible and there are a number of industries it could support (& vice versa). The problem is that no one of them makes it worth the initial, fairly large, investment on its own. However the tax free/low tax regime likely is a very good fit with financial services.
Note that Rapture ended really, REALLY badly.
Not that the technology element isn't interesting, but if the theory was workable, what would the, no doubt very impressed with themselves, 'traders' running this be adding to any part of humanity in any way shape or form?
Libertarian 'economics'... blech!
I suppose the one good thing is that at least a bunch of them would be shipped off to live in the middle of the Atlantic. :p
I find it pretty contemptible that all this time, energy, and technology is being wasted on ways for useless people to milk even MORE money out of the world economy without contributing a thing.
This is why we need a financial transactions tax.
How long until there is a demand for a fair playing field for trades where geographic proximity doesn't give a trader an advantage?
Imagine an exchange where trades are made at fifteen minute intervals with all trades registered in the last fourteen minutes executed as a single simultaneous transaction.
Fifteen minutes is just notional. The time interval used could be larger or smaller but all trades would be simultaneous and executed without regard to what order they were made. There is no logical reason to avoid allowing everyone to have a relatively flat playing field.
Never.
Because the stock exchange gets money for hosting and the high barrier for entry ensures that this benefit is only for the rich and powerful.
Rather than a transaction tax, a minimum holding period should be enforced and is pretty much a shoe-in with the technology: you must keep your stock for at least 24 hours before you can sell it on.
Even that doesn't allow much in the way of INFORMATION on which to base your decision to sell on, mind. But it's far more possible than microsecond trading, where all you're doing is shaving off for yourself any profit someone else's work is trying to gain.
Since the stock exchange already has to timestamp these transactions to the required accuracy, a timeout of 24 precise hours should be easy. To avoid leap-seconds etc, the figure could be stated in absolute seconds: 86400s.
Do you know what's even more idiotic about this trading? They did a computer model of the stock exchange and proved that the trading regime (at the time, seconds to minutes minimum turnaround) was inherently unstable. Because a computer can only be used to make the decision to buy in such a short time on so many products. Therefore one computer sees a sale (someone needs to cash in to pay inheritance tax, for example) then sells because the price is going down. Other computers see TWO transactions selling. They sell. The spiral begins.
For stock market traders, this is no problem: they get paid for each move. The more moves, the more they get paid. Even if it's a computer doing it.
24 hours is too long a holding period. An hour is perfectly fine.
I say this because I've wound up doing a buy and a subsequent sell on the same day. Not because I planned it that way, but because the market moved. All I did was buy, then, once the buy was complete, put in a sell for what I considered a reasonable profit, as a "good till canceled" order. As it turned out, in a few cases, the sell executed the same day, a few hours later.
Now, this didn't require any fancy technology and anyone who does online trading could do it. So why should it be prohibited?
(By the way, I have no problem with the idea of a transactions tax.)
"24 hours is too long a holding period. An hour is perfectly fine."
Why is 24 hours too long?
How long does it take to get reliable information rather than just a message to someone so that they can make a decision that involves, say, a billion dollars?
Do company CEOs have to made decisions in an hour? Do they fail if they take several days to come to a decision?
No.
An hour is enough to quieten down the inherent instability of the stock market (to an extent), but it DOES NOT fix the underlying problem: stock is bought as a gamble, not as an investment.
"I say this because I've wound up doing a buy and a subsequent sell on the same day. Not because I planned it that way, but because the market moved."
Did you know WHY it moved? Or did you just notice that it had?
You see in an hour you cannot both get the message "Market moved" and the information "Because the company is about to go bankrupt" (or whatever the reason was).
And you can't do stock market trading without fancy technology, so your lack of fancy technology is irrelevant to the problem. It's like saying your cart doesn't need the roads, but there's a sodding great big road between the two places you want to go.
All transactions are timestamped. If they aren't timestamped to the microsecond, you cannot do microsecond transactions. If they're not timestamped to the millisecond, you cannot do millisecond transactions. Therefore the timestamp cannot be used as a reason why a 24-hour minimum wait period cannot be used. The stock market already has to have a timestamp on transactions.