We can either elect Trump, or not-Trump as president. Everyone agrees that we want the price of beans to be cheaper. So there are 4 possibilities: 1) expensive beans + Trump president 2) expensive beans - Trump 3) cheap beans + Trump 4) cheap beans - Trump
An attacker doing what you describe would buy up many shares of (3) and (2). To try and convince us all that if trump wins, beans will be cheap. The attacker will need to keep buying up (3) and (2) to keep them more expensive than the alternatives. So the attacker is betting like $100 for every $50 others are betting.
Assuming the attacker has more money than everyone else who is willing to participate in this market put together, then the attacker could cause us to think that Trump will make beans cheap.
But, there is still the possibility that beans will be expensive, even though Trump won. If that is the case, then the attacker will lose all their money, and will not be able to afford to do attacks like this again.
Attacking a market like this necessarily means paying more for something than it is worth, which necessarily means that on average, you are losing a big chunk of the money that you use for manipulations. And everyone who bets against you, on average they are taking your money.
I wrote futarchy onto the blockchain with Amoveo.
Two responses. First, I'm thinking the manipulator has an outside incentive. If trump wins, they get a million dollar kickback. So losing 500k on bets is acceptable to them.
Second, I'm thinking of the following attack that actually loses nothing (if it works). The attacker doesn't bet at all in the market conditioned on trump being elected. She only bets in the other market. But she bets really heavily that beans will be expensive in that case. So we elect Trump. All trades in that other market are cancelled and the attacker gets all her money back. And she isn't even exposed to any risk on the bean price on this timeline, so to speak.
However, there is a profitable manipulation of this mechanism. A manipulator with deep pockets can bet heavily that some policies will have terrible outcomes. These policies are never selected, so the manipulator is never harmed by this misinformation.
I guess Hanson would reply that the rest of the market hopefully has deeper pockets than the manipulator and might be able to make money by correcting the predictions, but that sounds a bit fragile to me. An example research paper on the issue (pdf): http://yiling.seas.harvard.edu/wp-content/uploads/DM_full_ve...