Trading technology and the threat to IR
Anyone who is not at least mildly worried that computers can trade stocks faster than people can follow the trades is surely on the way to not knowing where their DNA ends and the silicon starts. Neurologists have calculated that the human ‘now’ is about 2.5 seconds long, and events we observe that happen faster than that generate a fuzzy chronology in our minds, so that it is difficult to remember the actual sequence of data inputs – or memories, as they used be known.
When the original joint-stock companies began their sales process, it was mediated at the speed of a fast horse or speedy ship, and a quill pen scratching out the details in copperplate established ownership. But when securities are whizzing round at close to light speed in virtual reality, who owns them at any given moment? Could a security be in a state of superposition like a quantum particle, bought and sold simultaneously?
Did I buy it back from myself before I had sold it to someone else? Is it insider trading if a security has the same owner several times in a nano-second, less time than ‘now’ for a human brain? If the intellectual foundation for the whole premise of the market is that fully informed buyers and sellers are trading to approximate the true value of a security, then what purpose is served by these lightning-fast quantum churnings? The mind-bogglingly tenuous connection between humans and finance is stretched even thinner with the news that the Associated Press now generates its reportage of most quarterly earnings filings without any human intervention. One of the pioneers estimates that the system, Automated Insights, can generate 2,000 articles per second, which is far faster than any human can actually read them.
For what it’s worth, AP says it was originally reporting on only 300 companies’ quarterlies, but now its computers are publishing 3,000. Why? In the real world, quarterlies are of dubious value anyway, boosting the trend toward short-termism. European companies can manage without them with no visibly significant effect on market efficiencies.
Surely, such prodigious financial literary output can only be meant for other computers to read – which is worrying. At least when humans are editing they can take into account that Airbus or Boeing billing for a plane the day before or the day after the reporting date could distort the picture. Computers can be programmed for some such contingencies, but they don’t have the sharp eye for the anomaly one would hope a human analyst would spot.
Markets reporting has indeed become so formulaic that its algorithmic prose already reads as if a bored computer generated it. [(Wall Street) OR (US investors) OR (the City) OR (markets)] [(rose) OR (fell) OR (met) OR (did not meet) expectations] on news of [(GDP) OR (oil prices) OR (employment figures) OR (interest rates)] AND (random-pick analyst) comments (rhubarb, rhubarb, rhubarb).
Until now, however, such reports have been like magical incantations designed to lend verisimilitude to an otherwise bald and unconvincing narrative, to lend the comfort of ritual to human actions. Humans can roll their eyes, and say, ‘So what?’ as so many journalists do when confronted with floods of ungrammatical, uneventful corporate flackery. But perhaps the hypertext is really on the wall: AP is now adding ‘forward-looking guidance’, and the implications are upsetting. Computers will now generate reports suggesting action, which other computers reading the reports will use to execute trades online.
Will the last investor relations officer to leave the building please ensure the computers are connected to a safe power outlet and switch off the coffee machine? Then rush down to the unemployment office before the rest of the world joins the line.