So to summarize, if a credible attacker announces his/her intention to destroy a POS currency, holders of that currency are likely to sell to the attacker very cheaply due to their belief that the attack will be successful and thus their coins will be worth nothing if they sell them. Very interesting point. In fact, it presents a "prisoner's dilemma" of sorts, where an attacker can be thwarted if the majority is able to resist the pressure, but where it's in the interests of an individual to break consensus and deal with the attacker.
6 pages and this is the argument:
"Then, there are two equilibria for our game. In the first one, when U > r(N + 1)/β, agent
0 buys strictly more than half of the coins and actually kills the CC. Since this is anticipated
by all the other agents, the latter are in competition to sell to agent 0 their coins, who they
know have already no value. The attack can be undertaken at no cost."
and its bad.
It doesn't count fiat as cost and equates 'known value' with price. Beyond that, it can be easily used for POW coins as well as it doesn't really matter what you are buying, hashpower or coins.
Incorrect. In PoW, the network is secured by hardware, so the owner of a majority of coins is irrelevant. To anticipate your argument: yes, if one person decides to buy the majority of coins and attempts to destroy them by "dumping" that will drive the price down and reduce the economic incentives to mine. EXCEPT, that as fewer people mine, difficulty adjusts and mining becomes less expensive relative to rewards. Not only that, but as that paper points out, the cost of operating mining rigs is marginal (especially with difficulty adjustments).