Any statement about improvement in liquidity resulting from HFT typically is based on some measurement of liquidity in terms of assets or volume of trading, without any assessment of its actual benefit.The stock market is enormously beneficial insofar as it allows firms to raise capital easily; but the vast majority of stock trades involve previously existing shares and thus do not have a direct effect on the firm whose shares are being traded. There is of course some correspondence between a firm’s real economic activity and the trading of its shares; for example, the share price affects whether a management buyout or hostile takeover is attractive. However, this correspondence is not affected to any significant degree by HFT: HFT activity just changes the prices slightly more quickly.
This new paper by Budish, Cramton, and Shim, which sparked this discussion at Marginal Revolution, proposes frequent batch auctions, i.e. trading intervals on the order of one second. One parameter in the paper is the cost of short delays in trading imposed by restricting HFT. The authors do not attempt to measure the actual cost, but they suggest that it is small. I don't see how this kind of cost is of any significance when the delay is on the order of a second. Does it capture the trader's impatience at having to wait to execute a trade? Lost interest while assets sit in limbo? Fear of missing some deadline? Even if the delay is minutes rather than seconds, what exactly is the problem? There could certainly be some problems if you have to wait a month, but once the market is thick enough that you can pretty much trade any time you want, what is the benefit of being able to trade at 12:00 instead of 12:01?
Budish et al point to the high costs of HFT infrastructure as well as greater stability in trading markets to justify the imposition of frequent batch auctions. I'm not convinced that there are any social benefits of HFT at all, especially in comparison to other forms of algorithmic trading (one of the points I made in my paper referenced above). If there are benefits, they are very small in comparison to the costs. I am no kind of finance guy, but am I missing something?