My suspicion (which I can't prove--this is just a gut feeling) is that there is an economically correct price for any given thing, and that markets will move toward that price. This economically correct price only changes when factors in the "real world" change, or we have reason to believe factors in the real world are going to change (e.g., bad weather might reduce crop yield, or forecasts of bad weather make is expect a reduction in yield).
Actors in the markets get information not only from real world sources (e.g., news reports), but also infer information from the activity of other actors in the market, so there is a bit of a feedback look in the market.
My suspicion is that as you increase the frequency of trades, starting from very infrequent (e.g., people trading in person directly with the person they are buying or selling from) and then going to faster and faster methods (trades by mail, then telegraph, and so on), there are two effects. First, the market is able to more quickly converge to near the "correct" price for the item, and there is some increase in fluctuation around the correct price due to strengthening of the feedback loops.
I suspect that there is some optimum trading speed, which depends on the rate that the real world events (things like weather changes, fashion, wars, and so on) happen, and when you push trading speed past that you stop gaining on convergence speed, but continue increasing instability due to the feedback loops.
The result is a market that is less and less tied to the real world. In effect, the market ends up makingup information from the trading activity itself. The trading essentially ends up being based on noise rather than signal. It's hard to see how that can be good.
trading activity is information because it is real people putting real money at risk and expressing their views of what the assets are worth. As in any competitive market, HFT traders cannot make money if they only trade with themselves - that is a mean zero proposition. But by trading with people who price the assets incorrectly at a point in time, they help move asset prices to their "correct " levels
Actors in the markets get information not only from real world sources (e.g., news reports), but also infer information from the activity of other actors in the market, so there is a bit of a feedback look in the market.
My suspicion is that as you increase the frequency of trades, starting from very infrequent (e.g., people trading in person directly with the person they are buying or selling from) and then going to faster and faster methods (trades by mail, then telegraph, and so on), there are two effects. First, the market is able to more quickly converge to near the "correct" price for the item, and there is some increase in fluctuation around the correct price due to strengthening of the feedback loops.
I suspect that there is some optimum trading speed, which depends on the rate that the real world events (things like weather changes, fashion, wars, and so on) happen, and when you push trading speed past that you stop gaining on convergence speed, but continue increasing instability due to the feedback loops.
The result is a market that is less and less tied to the real world. In effect, the market ends up making up information from the trading activity itself. The trading essentially ends up being based on noise rather than signal. It's hard to see how that can be good.