HONEY I SHRUNK THE MARGINS

Dick Pountain/16 August 2001/12:06/Idealog 85

The word 'snipe' has so far only had two meanings for me: a small game bird (very good roasted and eaten cold with a glass of claret for breakfast) or to shoot at someone from a place of concealment. However thanks to Davey Winder's excellent Online column in this issue I've just learned a third meaning, albeit a metaphorical derivative of the second one above. While describing his extensive use of the eBay auction site, Davey discusses the practice of 'sniping' which means winning auctions by using a special software utility that places bids at the very last second, too late for the previous highest bidder to respond. Fascinating in itself, this notion of sniping also triggered off some questions that I've been gestating for quite some time about the potential effect of electronic media on the economy as a whole.

In its standard form, the sniping tool that Davey uses allows you to set the timing for your snipe attack down to a limit of 10 seconds from the announced end of the auction, and it has proved so effective that he has only lost 4 out of 50 or so bids made this way. However putting on my games theorist hat for a second, it's obvious that such a strategy can only be effective as long as it remains semi-secret, that is, while not all the participants know about it. Sooner or later all the participants would suss what is going on and acquire a copy of the sniping tool themselves, so that eventually everyone would be sniping and all the bidding would happen in the final few seconds of the auction. The outcome of such an auction would then depend purely on technical factors such as what algorithm eBay uses for queueing bid requests, and no longer on the strategy of the players. This situation has not yet been reached, but Davey makes it clear that such a possibility exists, because the tool vendor also offers a premium service. The standard tool is freeware, but if you choose to pay you get a special version that allows sniping down to 3 seconds instead of 10 seconds, and gives you a permanent advantage - and because it costs money, this version will always be less heavily subscribed than the free one.

What this points up is that computers can profoundly affect the way that markets work. Over the centuries markets, for example stock exchanges, have evolved to a state of considerable refinement based on game-playing between human participants, carried out over human time scales of days, hours and minutes, with their outcomes in some way reflecting the human psychology of the participants. Computers can overturn this situation completely, reducing each market transaction to a computation performed in milliseconds, with no psychological constraints and with total efficiency.

Futurologists like Nicholas Negroponte have long predicted the coming of a society in which all goods and services will be auctioned online and each citizen will have software agents that purchase these goods and services for them at the best possible price. Something very like a miniature model of this situation already exists in the New York stock exchange where automatic program trading is common and safety trips have to be built in to short-circuit massive panics. Going further still, it's always been a temptation for utopian thinkers who don't like money very much to imagine that technology can remove the need for it altogether. If your salary is paid directly into your bank account and you spend most of your income using a credit card, then in principle you are already participating in such a cashless, digitally-controlled goods flow - and it might seem only a small and logical step from there to a society in which income, taxes, benefits and purchases are all consolidated and virtualized. Politics would then be reduced to an argument about what algorithms to use for the distribution of goods: hard-core market devotees believe that markets provide the best algorithm, always tending toward equilibrium and maximum efficiency, while hard-core state socialists believed (wrongly) that they could write their own algorithms, in 5-year chunks.

Such an automated economy would be one huge dynamic system in which virtual value flowed around the world and regulated the apportioning of goods and service to people. However we know a lot more about such systems than we did 20 years ago, and one of the things we know is that they are rather less stable and predictable than classical market theory believed. We know about stuff like catastrophe theory, chaotic systems and algorithmic complexity; we know that the very same dynamic system may settle down to one of several different equilibrium configurations, or, for certain values of the variables, enter a wildly thrashing chaotic state (that's why the New York stock exchange needs those safety trips); we also know some basic principles, like the fact that reducing the amount of buffering in a dynamic system, by reducing delays and reservoirs, makes the system more sensitive to changes of input and more likely to fluctuate wildly.

What I don't know, and would dearly love some economist to explain to me in plain language, is what overall effect such automation of markets and trading is likely to have on profit margins. It seems intuitively (though intuition is often wrong in these matters) that it must reduce them to almost zero. For example the recent fuss in the UK about 'rip-off Britain' revolved around imperfect information: when most of us didn't really know that cars were much cheaper in Belgium we happily paid through the nose, but now we do know the car vendors are taking a beating.