There are now going to be ten nominees for the best picture category.  This poses a danger with  choosing the film that gets the most votes as the “best picture”: what if it is the worst film according to people who put other movies at the top?  Maybe they all agree on the second-best movie and if you take this into account, you’d have a quite different “best picture” winner.  To take this kind of thing into account, the Academy of Motion Pictures has come up with a new scheme:

Voters will be asked to rank the 10 best picture nominees in order of preference, one through 10. Davis says that the category will be listed on a special section of the Oscar ballot, detachable from the rest so that a separate team of PricewaterhouseCoopers staffers can undertake the more complicated tabulation process.

Initially, PwC will separate the ballots into 10 stacks, based on the top choice on each voter’s ballot. If one nominee has more than 50 percent of the vote (unlikely, but conceivable some years), we have a winner.

But if no film has a majority, then the film ranked first on the fewest number of ballots will be eliminated.  Its ballots will then be redistributed into the remaining piles, based on whichever film is ranked second on those ballots.

If those second-place votes are enough to push one of the other nominees over the 50 percent threshold, the count ends. If not, the smallest of the nine remaining piles is likewise redistributed. Then the smallest of the eight piles, then the smallest of the seven…

Eventually, one film will wind up with more than 50 percent.

Which is the voting system that is more manipulable via strategic voting, the original one or the new one?  How is it manipulable?  Harvey Weinstein will be willing to pay for advice.

Update: As Mallesh says below and David Austen-Smith pointed out to me in an email, this voting system is the Single Transferable Vote system.   It violates Arrow’s independence of irrelevant alternatives and is manipulable as the Gibbard-Satterthwaite Theorem applies when IIA fails.  I guess manipulability per se is not the right criterion.   One should study the (Bayesian?) Nash equilibria of the voting system in question and compare them via some notion of welfare to the equilibria of the original system.  The original system (as far as I can tell – I am not an Oscar voter), was simply one man-one vote with the movie getting the most votes picked as the winner.  The new system is much more complicated as you have to report a whole ranking.