RCP gives Clinton a 3.2% advantage in the popular vote and a 301-237 edge in the electoral college. 3.2% nationally is a healthy margin, but RCP was only off by that much in 2012, to Obama's favor. If you drill down on a state-by-state level, it is much less clear. Two states, FL and NH, which would deliver Trump a victory are within 0.2% and 0.6% chance. State polls are routinely off by 2% or more.
The below table is a comparison of final state polls in 2008 & 2012 versus their actual. I only examined polls where the lead was less than 5%, as those states would have more frequent polling. Polling is a lagging indicator. National polls are more frequent, so state polls can be supplemented with national trends. Both times, national polls showed late momentum for Obama and the results broke for Obama. Last nationals polls have broke slightly for Clinton.
The results are uncertain but I think Trump is in a better spot than Romney was. For Romney to have won in 2012, he would have had to pickup five states, three of which Obama was leading by 2% or more. Trump has to win two states where Clinton is ahead by 0.6% or less. The outlook is certainly worthy of 31.5% chance of winning that FiveThirtyEight gives Trump.
One misconception is that margin of error is absolute - i.e., if a poll has a 4% margin of error, and the poll shows candidate A down by 3.7%, many pundits will say that is a complete tossup. The 4% likely represents three standard deviations, so a 3.7% deficit really means about a 5% chance of winning, not a 50/50. Being down by 0.2%, however, is very close to 50/50.
2016 Data
For comparison, Electoral-vote.com gives NC and ME#2 to Clinton for a 317-221 edge.
2012
Realclearpolitics was the least accurate for the popular vote, being off by 3.2%. For the electoral college, they only called Florida wrong, which they had a 1.5% margin for Romney, but it went for Obama by 0.8%. Virgina was correctly called, though they called a 0.3% margin for Obama when the actual margin was 3.8%. Electoral-vote.com had the same for Obama (303), but put NC as a tossup.
2008
The 2008 race was quite accurate: They were off 0.3% on the popular vote in Obama's favor, and put two states, Indiana and North Carolina, in the McCain column, when they ultimately went to Obama. McCain lead Indiana by 1.4% and went for Obama by 1.1%. McCain lead North Carolina by 0.4% and it went to Obama by 0.3%.
2004
The 2004 race, the popular vote err'd in Kerry's favor by 0.9%. Realclearpolitics didn't have a electoral state map which I was aware of, but I was following electoral-vote.com at the time which gave the edge to Kerry. The err'd on Iowa, New Mexico, Wisconsin, and completely missed the mark on New Jersey.
No comments:
Post a Comment