Electoral College or Popular Vote?

[Mr. Morden]

In a close election, allocating the EVs proportionally by vote within each state actually tends to tilt things slightly towards the Republicans (relative to their showing in the popular vote), since they tend to do a bit better in smaller states, which have a bigger proportional bonus in the EC with the +2 EVs for senators.

For example, I think if you allocated the 2000 EVs by popular vote in each state, then Bush wins the electoral college more decisively (without the need for a recount), despite losing the popular vote nationwide.

[Mr. Morden]

I think the debate is breaking down over the semantics of what it means for a vote to count.  The way that I would put it is that under the EC, the votes of both Democratic and Republican voters in Texas do in fact count, in that they determine how the electoral votes of Texas go.  However, these voters lack "electoral leverage", which means that their votes are determining an event for which we're already ~95% sure of the outcome, since we're pretty sure that the Republican candidate is going to win Texas.

Interestingly, electoral leverage is based entirely on our ignorance of how the vote is going to turn out.  The states where we pretty much know who's going to win lack leverage, and the candidates ignore them, whereas the states where we're ignorant of who's going to win beforehand are the states that the candidates care about.

One could imagine a thought experiment in which the electoral college is left intact, but the state boundaries are all redrawn at random, all records of past voting records by precinct are destroyed, and statewide polling is banned.  The candidates would be forced to campaign nationwide, and the campaign would play out as if it was determined by the popular vote, because no one would know which voters were living in states that would end up being competitive and which were living in states where it would be a landslide.  The ignorance of whose votes "mattered" in such a system would make it "more democratic".

[Mr. Morden]

In a close election, allocating the EVs proportionally by vote within each state actually tends to tilt things slightly towards the Republicans (relative to their showing in the popular vote), since they tend to do a bit better in smaller states, which have a bigger proportional bonus in the EC with the +2 EVs for senators.

For example, I think if you allocated the 2000 EVs by popular vote in each state, then Bush wins the electoral college more decisively (without the need for a recount), despite losing the popular vote nationwide.

Wrong. While you're right Bush would get more EVs, he only gets a plurality. By one. And it's thrown to the House:

Ah, you're right on that.  Sorry, I was getting it mixed up with the scenario where the EVs are allocated by CD.  In that case, I think Bush wins.

[Mr. Morden]

I wonder what the dominant strategy would be for the parties in a proportional situation?  3 EV states would be pretty irrelevant unless they are really close.  You get 2 EV with 51% but need to get to 84% to get the 3rd one.  However, in CA every 1.8% of the popular vote gets you another EV.  Large state's like Ohio and Pennsylvania are basically never going to be outside of 55/45 which means they will always split evenly or within one vote of even in an election that is remotely close.  So it looks to me like it would be all about turnout in the large states.  D's would move their swing state turnout machine to CA and NY to try to get >70% and R's would do the same in TX.

It would be like Obama's 2008 presidential primary campaign strategy, except that was fought at the CD level.  The Dems use proportional representation by individual CDs for their presidential primaries, so the Obama campaign targeted the individual CDs where they thought the race would be close to a threshold for getting an additional delegate.

In the scenario you're suggesting, with proportional EVs by state, it would be similar.  The campaigns would poll nearly all 50 states (might not be worth it in some of the 3 or 4 EV states), and try to work out which states are going to have popular vote %ages that put them near the dividing line on getting an additional electoral vote.

So, for example, if you have two states with 8 EVs each, and one of them has polls showing an even ~50/50% race, while the other one shows you at about 55% of the vote, then you'd put money into the latter state and ignore the first one, since a ~50/50 state will split its EVs 4-4, whereas getting up to 56.25% of the vote gives you a 5-3 split.

You might say that it would be more cost effective to put your money in the biggest states, because a smaller %age shift in the vote in those states is needed to get another electoral vote.  But bigger states are also more expensive to advertise in, in order to get the same movement in the polls.  You're right, though, that under this system, a larger share of the states that you completely write off would be the small states.  It would generally pay off to spend at least some money in each of the largest states, since most of them would probably be close to the threshold of giving you an extra EV.  And it wouldn't matter if it was one of your base states or not.  Both parties would be spending $ in both CA and TX in order to try to get an extra EV.