There are two ways this can be done. Neither is in the interest of the state to do so.
1. A True Proportional Allocation of Electoral Votes
Scenario: Electoral votes are apportioned according to the popular vote, with the winner getting the "round up."
e.g. A state with 10 electoral votes splits 53% - 47%. The winner gets 6, the loser gets 4. (Alternatively, you could go 5-5, but this makes it even worse!)
Colorado came the closest to doing this back in 2004, but abandoned it for a pretty good reason - going to proportional representation in electoral college votes almost guarantees that your state will be ignored.
The article even says as much:
"For the first state to do this unilaterally, we go way down in importance," said Dave Kopel, research director at the conservative Independence Institute in Golden. "Because there's not much to gain ... a candidate would be nuts to spend time in Colorado."
Imagine, for example, a state with 10 electoral votes, apportioned according to the popular vote.
Now, let's take the case of Alaska, which, according Nate Silver, saw the most dramatic swing from the 2008 election to 2012. (In the case of Alaska, the reason is pretty obvious - Sarah Palin's involvement in the election took a normally solidly republican state, and made it extremely Republican.) This most dramatic swing of the state represented a change of 28%.
In our hypothetical state with 10 electoral votes, this would represent a swing of 3 electoral votes - maybe going from say 9-1 to 6-4. That represents the top end of a state. Putting as much time and effort into the state yields, at most, 3 votes. If any other state doesn't do this, that state is worth 10 votes, just for swaying a few people. From an effort perspective, that state isn't worth campaigining in.
Now, again, Alaska in reality, only has 3 electoral votes. Thus, even with such a dramatic swing, 28% is less than 1 elector. As such, just about every 3 vote state will always split 2-1. And, in just about every one of the 3-vote states, it almost always goes in the same direction.
And, just to add insult to injury, remember that a blowout election is 60 - 40. Outside of the People's Republic of Washington, DC, getting more than 65% for any one party is well nigh impossible. So, again, in our hypothetical 10 vote state - what does a blow out get you? One, maybe, just maybe two votes. Again, compared to an all-or-nothing state, your votes just aren't worth that much.
In 2012, Obama won 26 states, and as such, would have been the winner, albeit by a much smaller majority in the Electoral College.
With such a minor return for such a major investment, a candidate has thus very, very little incentive to expend effort in such a state. It would be as if we went to a strict popular vote - why then would candidates bother with rural voters when courting their concentrated city cousins has much more result for the effort.
2. Congressional District Apportionment
Scenario: Winner of each Congressional District gets 1 electoral vote, and the winner of the state gets the remaining two electoral votes
e.g. Maine & Nebraska
Note: I don't consider this to be proportional, but apparently others do. As such, I'm addressing it.
Overall, the commentary above holds true - going to such a scenario makes any effort in the state less desirable. Unlike the "9 battleground states" of 2012, you get the "35" (of 435) competitive house districts. Again, as Nate Silver shows, the number of congressional districts where the opposition party has a real chance has been narrowing dramatically. In 1992, there were 103 seats that could have swung either way. In 2012, there were 35.
Safe districts (for better or worse as @DjClayworth points out) simply don't get the attention. Likewise, one of the defects of a popular vote system (corrected by the Electoral College) is that rural votes don't get a lot of attention for the same reason. (If you have a choice of reaching 100,000 voters by buying ads in one TV market or ten, where are you going to buy?)
In practice, neither Maine nor Nebraska saw a lot of activity during the 2012 campaign. And, even though Obama won NB 2 in 2008, his campaign was unwilling to expend major resources there in 2012 - again ignoring it.
Looking at the results of 2012, assuming that the popular vote was unchanged, but that every congressional district voted according to the party of its member of Congress (not entirely valid, but the best proxy I could come up with)
Obama would have lost, 256 - 282
The methodology is as follows:
- Assign 2 electoral votes for the winner of the popular vote in each state - Obama 52, Romney 48
- Assign each Congressional District according to its winner - Obama 200, Romney 234.
- Assign all three votes for DC to Obama
- Assign Jesse Jackson, Jr's vacant seat to Obama
Romney = 48+234 = 282
Obama = 52+200+3+1 = 256
There are 9 states where the Presidential pick differs from the congressional delegation:
- Colorado
- Florida
- Iowa (tied 2 – 2)
- Michigan
- Nevada (tied 2-2)
- New Jersey (tied 6 – 6)
- Ohio (12 – 4)
- Virginia
- Wisconsin
Interestingly, this exactly corresponds to the "battleground states" of 2012.
Update
Using the Daily Kos analysis of presidential preference by Congressional District, we can get actual 2012 results:
- Obama won 205 congressional districts, Romney 230
- Again, allocating 52 state popular winner votes to Obama, 48 to Romney
- And, allocate DC's 3 to Obama, we get...
Obama loses to Romney 260 - 278
Interestingly, that is within 8 votes of being the exact opposite result as 2012.
Obama still would have beat McCain 291 - 247, however, in 2008.
tld;dr>
The point here is pretty simple - going to a proportional system guarantees that your state is either no longer in play or not really worth the effort. (So you push 1 vote in a state. Big deal!) Because of this, going to such a system simply reduces the impact of a state in an election, and is therefore not in the interest of the state to do.
A ME/NB result looks to favor Republicans, however. This is due to the fact that rural voters tend to be more conservative, and are perhaps, over-represented strictly based on population.
