In terms of unrealistic assumptions in Robin Hanson’s series of papers on manipulation, the major ones have been out there since at least 2004.
Despite some limited evidence, the insistence on traders needing to know the direction of manipulation isn’t too compelling since the direction will be manifest insofar as the price is “wrong.” “Noise trader” is a politically loaded and misleading term. Misleading because it suggests that the mean effect will be zero, when in reality “noise trading” usually takes the form of feedback trading. Lack of feedback trading is a significant assumption in the Hanson manipulation papers. Fortunately, prediction markets have objective settlements at specified times, unlike traditional assets where the meaning of prices is open to interpretation, making them more prone to feedback trading and irrational booms and busts.
With prediction markets, conditions for manipulation are more favorable when the settlement is far off in time, and when there are subjective inputs to the settlement, e.g. in politics. A distant settlement simultaneously makes it less clear what the real price should be, and delays manipulator losses because there is less incentive to correct price. At the limit, a manipulator could introduce a price distortion when a contract is launched, only to reverse position for small liquidity-related loss immediately before settlement, thereby destroying the markets “integral” of error over time.
Another big assumption, also identified by Paul Hewitt, is that traders have equal account sizes. But maybe this isn’t a huge problem if settlement is forthcoming, and maybe the issue could be mitigated with additional exchange disclosures, such as the standard deviation of position sizes in a given market. While this could discourage liquidity as large traders would become paranoid about their positions, it is essentially a “soft” position limit, and traders would be forewarned of one-sided markets (which could of course be the result of someone well-informed, but I – the google-anonymous* writer – would bet that more concentration comes with more error on average… this can be tested by someone with the data, of course maintaining trader anonymity)
Even accounting for long-term settlement, feedback trading, semi-subjective settlements, and account size imbalances, it seems one would have to abide to an overly rigid tenet of “do no harm” to hold that prediction markets are, on net, a bad idea. (Do no harm is of course abhorrent to libertarians, and even doctors don’t actually follow such a rule.) Moreover, some pathologies like political self-fulfilling prophecy will only happen if prediction markets have already demonstrated their value and have become more popular. But even if one believes in their long term success, single pathologies can damage one’s reputation permanently… if one plans to die at a reasonable age.
[*Given the political climate, many firms have issued directives to employees to not engage in even the slightest appearance of impropriety, which might include blogging on manipulation.]
What exactly do you have in mind as “feedback trading”? My theory paper does *not* assume traders know the direction of manipulator trades, and there are no account sizes in the model.
Feedback trading just means the kind of momentum trading that is pervasive in traditional assets, again, less so in prediction markets. In the biastest experiment, traders were given formal “clues” about the settlement, but for many market participants, the best “clue” (even rationally, if lazy) is recent price action. Even if feedback trading was possible within the experiment, the outcome (manipulation attempts were corrected) suggests that it wasn’t prevalent.
In this experiment, traders were given equal endowments of shares/currency… i.e. initially had equal account sizes. Yes, there were an equal number of manipulators and non-manipulators, but they could not coordinate. Even if they were implicitly coordinating, this is not the same as a single large trader influencing the market. Yes, in the theory paper trading sizes were variable, but according to the same parameter for each trader. Maybe if there were a large supply of potential traders able to frictionlessly join the manipulated market, a manipulator’s relatively deep pockets wouldn’t matter.