During the course of my PhD, and into my postdoc, I have performed "first principles" molecular dynamics simulations of various molecules and materials. What's the connection with high frequency trading? Well, those calculations, which involve a pretty decent solution to the SchrÃdinger equation, typically involved simulating the motion of atoms over timescales on the order of picoseconds. One of the advantages of such simulations (compared to say classical molecular dynamics) is that there are no empirical parameters involved. This is physics speak for "we didn't force the answer we wanted with a fit". One of the disadvantages of these methods, however, is that they can only look at processes which occur on short timescales, due to the high computational cost of the method. Just to be clear - LOTS of important physical processes occur on timescales longer than picoseconds. I would have thought the economy was one of those processes.
Now, back to high frequency trading. Why the push to faster transactions? The answer is simple: if you get there first, you do better. Believe it or not, firms with faster computers which are physically closer to the exchange have an advantage over those which are slower or farther away. The speed of light is actually something people consider when they are running financial transactions. If this seems ridiculous to you, it should. Traders argue that the existence of high frequency trading increases market liquidity. Let's assume for the moment that this is true (it's not 100% clear to me that it is) - is there a better way to achieve this goal? Everything I have read on this topic suggests that the typical business model is for large financial institutions to buy really expensive hardware which they place between legitimate buyers and sellers, siphoning off some cash in the process.
So, what is the alternative? How do we prevent time, money, and effort going into executing financial transactions on such ludicrous timescales?
I think the answer (as a Canadian I feel qualified to propose this), is that we need to establish Standard Financial Time. The concept is similar to Universal Time - if you want to build a railway, people need to agree what time it is. With SFT, exchanges need to establish the equivalent of a financial pulse. Transactions take place at regular intervals. The fact that you submitted your buy slightly before mine should mean nothing, provided they both arrive before the next beat. What should this timescale be? Well, if we want to have an efficient system (in the global sense), then everyone who would like to make a buy or sell offer needs to have an opportunity to do so. That means the pulse can be, at minimum, the time it takes to send a buy/sell order from any terminal in the world. For light to go from one side of the world to the other takes ~ 0.13 seconds. Anything faster than that, and you end up with financial types buying up computers in random spots on the earth trying to take advantage of information asymmetry due to the speed of light.
One problem this does create is that it is possible now to have two buys for a single sell, or vice versa. Fortunately the fix is straightforward. You either match the buyer and seller by a random number or, if you want to get fancy, you do something like the single transferable vote. Every buy order consists of an initial offer and a max offer (think of what happens when you bid on ebay). This works out well for sellers since they will maximize their sale price, and it works for buyers because they can minimize their payout. No matter how you go about doing it, the monetary advantage to shorter ethernet cables goes away; that can only be a good thing.
