That’s a common criticism of high-frequency traders, or for that matter of the older specialist system, or of some other trading practices; many MR readers make that point in these comments.
But is that a problem with uninternalized, Pareto-relevant externalities?
Let’s say many men visit a bar for its beautiful women. That said, if the women, on a given evening, don’t see enough desirable men, they go home. The beautiful women run away, but the plain women stick in the other bar, across the street, no matter what. Men choose between bars, knowing that the one bar will often be very crowded, but sometimes empty.
The negative externality from the beautiful women, if there is one at all, is on the bar with the plain women. That bar might find it harder to achieve critical mass and get off the ground. But the first bar is not made worse off by the possibility that beautiful women might show up or instead might “flake.” Men internalize that knowledge when deciding whether or not to visit that bar.
In other words, potential negative externalities from HFT (or other culprit trading strategies) are on the markets where HFT is not very active. Yet the complaints you hear are about the markets where HFT is very active, and those complaints don’t correspond to the theoretical argument which might make sense. They are wishing for the beautiful woman who sticks around at the bar no matter what.
The high-frequency traders are like the beautiful women. If their biggest “threat” is to stay home, we are not worse off for their existence. If we fear their flaky departures enough, we may prefer to trade in other markets or at other time horizons, namely very long.
All this is assuming that such flaky departures occur — relative to the relevant alternatives — and that point can be debated.