They most definitely can all be wrong, though.
Reminder to my fellow Romneyites: sign up to make phone calls - I'm calling voters in Nevada this weekend. It's lots of fun! Sign up! Act! Get more Romney voters to the polls!
Not many changes, because many of the pollsters are pretty quiet on the weekends for changes. I will be doing this up til Tuesday morning, and then Wednesday morning, I will compare these against what we know at that point. I have got an all-purpose excuse ready for any wrong predictions, which can be used by anybody. But I'm not going to unveil it quite yet. (Oh, I see Nate Silver has got an inkling of what that excuse will be)
First up, popular vote polling:
Reminder on Unskewed Polls outlier: that's their election day prediction, the other points are poll averages (their "unskewed" poll). I'm keeping the outlier to check against what happens.
RCP: Obama, 47.4%; Romney, 47.2% (barely moved)
Nate Silver: Obama, 50.6%; Romney, 48.3% (barely changed)
Unskewed polls: Obama, 45.8%; Romney, 50.0% (unchanged)
Electoral College projections:
RCP: Obama, 201; Romney, 191; Toss-up, 146 (no change)
RCP No Tossup: Obama, 290; Romney, 248 (no change)
Nate Silver: Obama, 306.9 Romney, 231.1 (shift of 1-ish vote to Obama)
Unskewed polls: Obama, 179; Romney, 359 (no change for ultimate projection)
Win probabilities: (will not necessarily add to 1)
Nate Silver: Obama, 85.1%; Romney 14.9% (1-ish percentage points to Obama)
InTrade: Obama, 66.5%; Romney, 33.6% (1 percentage point to Romney)
A few items:
Just one note on the state-by-state predictions Silver does, which I listed some
key states in the Nov 1 post, it's no big deal if he gets say, one "low probability" state wrong. But if quite a few of the "low probability" events occur, then he does have a lot of splainin' to do. (Hmmm, kind of like that whole CDO thing, some to think of it)
Now, getting things wrong doesn't mean he'd lose his job at the NYT. After all, they've kept Krugman for years.
For convenience, I'm going to list all his key tipping point states and his probability of Romney win in each of these:
Nevada: 10.5%
Colorado: 31.9%
Iowa: 16.9%
Wisconsin: 5.8%
Ohio: 15.0%
Pennsylvania: 2.7%
Virginia: 29.0%
New Hampshire: 20.6%
Florida: 53.8%
Let me remove the Florida data point, because I want to make a point of probability of what happens when you've got a bunch of low probability events and you wonder how many will occur. To simplify the situation, I'm going to assume all of the above are independent events (supposedly, Silver includes some correlations… or maybe he doesn't… I don't care either way. I'm doing the simple thing.)
According to Silver's probabilities, the chance of Romney getting all 8 states is so low as to be practically zero.
However, the probability of Obama getting all of those 8 states is only 22.25%. Which means that there's a 77.75% chance that Romney gets at least one of those.
I did the full distribution of the outcomes of number of states Romney gets from those 8, assuming Silver's probabilities are correct:
So, as I say, getting a few wrong isn't that troublesome for Silver. It's if he got a lot of these "low probability" states wrong. Maybe he's okay to up to 3 of those states going to Romney. Beyond that… yeah.
Here's the spreadsheet backing up the above. I did it in Excel originally, as google spreadsheets doesn't have all the functions that Excel does. So it pasted as a bunch of numbers and you'll have to check these separately.
One comment
Toggle Comments