HFT as Crazy Huge Market Failure

Financial traders are rewarded for trading on info that implies a non-zero difference between their expectation of the future price, and the current price. (Well, the “current” price here is the price their order would get if sent in now, and for positive revenue this difference must be larger than the info of an average trade, plus transaction costs.) If this difference is positive, they buy; if it’s negative, they sell.

After a trade, this trader info is incorporated into the market price, which helps everyone who considers that price in their decision-making. For example, more accurate stock prices help investors to pick firm activity levels.

The financial revenue gained from trading on info depends on the size of this price difference, but on little else. However, both the social value of such a trade, and its opportunity cost to a trader, depend greatly on the “duration” of each piece of info.

You can only profit from by telling a market about something that it will eventually learn in other ways, and it is too late for this once many traders know that many others know. Thus one can only profit by trading on info before its a many-know-it deadline, and you can only safely free up the assets you used in a trade, by reversing that trade, after this deadline. Thus the resource opportunity cost of trade is roughly proportional to its duration.

(Of course there can be many other kinds of costs to support a trading strategy. For example, risk costs may be larger when assets are held for longer.)

However, the social value of a trade is only gained for decisions made during the time period when the trade info adds to price accuracy. Which is the info duration. Thus this social value of a trade is also roughly proportional to its info duration.

Furthermore, this duration-til-deadline varies greatly. For example, “slow” info on future product fashions, or the success of innovation projects, may take years or decades to be revealed. In contrast, ~20% of trades are by “high frequency traders” (HFT), who typically trade on very “fast” info re prices in other markets. The deadline for the fastest HFT to arrive at a market with such other-market info is roughly when the second-faster HFT arrives. This is typically ~20-200ns for other markets at the same site, and ~50-500us for different sites (source: expert told me).

Thus five-year duration “slow” info is a factor of a trillion to quadrillion times slower than HFT “fast” info. So when slow and fast info imply the same price change, and thus give the same private revenue per trade, both the resource opportunity cost to trade and the social value are this much lower for a fast info trade, relative to a slow info trade.

When resource opportunity costs dominate costs, traders prefer faster info by this factor. But society prefers slower trades by this same factor. Together this suggests a total distortion that goes as the square of this factor. Which in the case of five year slow versus HFT is a factor of 1024 to 1030!

And this only considers the issue of info duration; other issues can only increase the ratio between the social value of info and its cost to traders. Thus familiar financial markets induce a huge distortion in trader efforts. In particular they induce traders prefer to seek out and trade on fast info, relative to slow info. The future gets neglected, due to an obsession with the next few milliseconds.

Now just because existing markets are failing here, that doesn’t mean the only solution is regulation. In my next post, I’ll discuss how different market structures can mitigate this problem.