I would certainly favor IRV for president and governor (and senator, under the current elect-one-at-a-time format)--preferably without the electoral college in the case of the presidency. However, it is actually more distortionary in some respects: It gives full power (of whatever office is being elected) to a candidate who was not the majority favorite. If adopted in the UK, for example, the result--at least initially--would be to essentially entrench the Labour Party in power (as Liberal voters would overwhelmingly give it their second choices).

In the 1992 US presidential election, IRV (with or without the electoral college) would have given Clinton a stronger "mandate" without giving the Perot voters who had made that possible any means to intervene between elections to hold him to the conditional nature of that mandate.

Now imagine if Perot voters had won 20% of the seats in Congress (with or without a share of cabinet seats), and you get a very different picture indeed!

To the commentator to this post before me, I would note that there are danger to jumping to conclusions about research results based on "just looking at the abstract." In any case, the claim by opponents of PR has always been that small parties have "disproportionate" influence in coalitions, period. In other words, that small parties get to "dictate terms" to the larger partners. The argument has never been based on closeness of result between the top two parties, but rather on the mere fact of one large party not being allowed to govern alone.

As I noted in my post, any situation in which the largest party gets a majority of seats (whether or not it also has a majority of votes, which it usually does not), the largest party by definition is given disproportionate power (i.e. all of it, even though it represents only a part of society!)

Coalitions resolve this problem by not putting one party in full power. There is little evidence that small parties in coalitions get to "dictate terms" and thus exercise power outside of their actual voter support.

Where they might have such "disproprtionate" power is in very close elections. The point in the paper (and abstract) that the comment latches on to is simply indicating that this outcome--close elections increasing the "pivotalness" of smaller parties--is not unique to PR systems.

"Small parties only have disproportionate bargaining power in elections where the two large parties are essentially tied. Given that this situation will produce a hung parliament and give small parties the balance of power in most electoral systems"

Well duh. Its only when their individual votes are needed to make a majority that the small parties get excessive power. Sure, if they aren't needed to make a majority, they don't have much power. But if they can work with either party and their few votes make the difference, they can dictate terms to a great degree. And if they are needed for a coalition government, then they can bring the whole house of cards down at almost any time.

While I can agree that there can be improvements in our electoral system, proportional electoral systems wouldn't be my first choice. Instant run off voting would be a much better tweak to the system. Not happy with the party and want to send a protest vote for the green or libertarian party? Now you know the odds are very good that the vote for the third party is less likely to give them the election and more likely to deny the election to your second choice. So you can pull the level for one of the two main parties or throw your vote away. But with instant run-off voting, you can vote for the third party and still see your vote count.

Such a system would be far less disruptive of the electoral system while still giving the potential to see changes. At the presidential election level in 1992, every EV but Arkansas and DC would have been decided by the IRV method. In 1996, 13 states for Clinton (149 EV) and 13 states for Dole (131 EV) would have been decided by IRV. That is enough votes to get beyond the magic 270 votes decided by IRV. In 2000 6 states for Gore (44 EV) and 4 states for Bush (54 EV) would have been decided by IRV, potentially changing the outcome. In 2004 two states for Bush (12 EV) and one state for Kerry (10 EV) would have been decided by IRV, though they wouldn't have changed the election since Bush won by more than 12 EV.

So three out of the last four elections would have been decided by IRV, potentially changing their outcome.

Of course, the third parties might start seeing their chances rise as people could vote for them without throwing their vote away or tipping the election to the party they like least.