Wednesday, November 7, 2012

Post Mortem: No Drama Obama, Nate and Me (November 7, 2012)

It wasn’t as close as any channel covering the election made you think.  The fact that the Presidential election was called so early, at about 11:15 PM EST by all the channels (and they are not allowed to call it before 11:00) is testament enough to the lack of real drama.  I had no doubt at all that Obama was going to win, but if I needed any early clue, it was when NBC called Michigan, Pennsylvania and WisconsinWisconsin! – between 9:00 and 9:30 PM EST.  Three good-sized states that had at one time been (or still were) battlegrounds, and I expected each to be relatively close, a 4 to 5 point margin.  But they were called fast, and their ultimate margin was 5 to 9.  Plus Florida was obviously very, very close from the outset – and that was Romney’s best swing state.  It was really over quite early.  “No Drama Obama” once again…the only drama was manufactured by the coverage.

I’m convinced that pundits view polls with great suspicion, and as a threat to their intuitive so-called genius.  Any number of individual polls can be wrong, of course, and are….the granddaddy of them all, Gallup, had Romney up by 5-6 points just before Sandy hit, which was 4-8 points different from virtually every other poll.  But what Nate Silver is demonstrating is that when you add polls together in some fashion – any fashion -- the margin of error goes way down.  If you sample 1,000 likely voters to represent the 120 million ultimate voters, the margin of error is about +/- 3.  But if you add ten polls together, effectively you are sampling 10,000 people, and the margin of error is +/- 1.

Pundits don’t get this.  I heard even the venerable Tom Brokaw complaining about the number of polls.  But the more polls, the better the forecast.  And this one was called exactly right.  Obama was going to win, if you looked at the polls correctly and believed that Obama’s “ground game” was as good as or better than Romney’s.  And most people believed that.  As did I.

So Nate Silver is today’s hero.  He’s been my hero for awhile!  But, to be a bit immodest, my calls were about 99.9% as good as Nate’s:

  • We both correctly picked 49 out of 49 states (plus the District of Columbia).  Florida remains uncalled; I predicted Romney would win by a nose and I will likely be wrong…Nate predicted a toss-up, which I guess is a bit more correct.
  • We both picked 31 out of 33 Senate races, and got the same two wrong, in fact the closest two races and the last two that were called.  North Dakota and Montana, where the Democrats registered razor-thin upsets, were both called in the afternoon.
  • In predicting the percentage vote in the swing states and close Senate races, Nate’s calls were ever-so-slightly better than mine.  Ohio’s Senate race is an excellent (and representative) example….he said Brown would beat Mandel 52-47, I said 53-47, and the actual was 50-45.  Our differences were even less pronounced in the Presidential swing states
  • And Nate did better on the Presidential popular vote:  he said Obama would win 50.8% to 48.3%, while I said 49.8% to 48.7%, and the actual (thus far) is 50.3% to 48.0%.
  • Nate did not pick the House races…there are still 8 that are not called; thus far I have picked 410 out of 427.
Nate’s methodology is more sophisticated than mine, a regression-based model that weights different polls in different ways.  I do more or less straight line averages, with the key decision being at what point do you consider a poll to be outdated.  I’m sure Nate’s slightly better precision is a credit to his sophistication, but the differences were immaterial and barely discernible.

My crude model for when the race would be called turned out to be too conservative….instead of Wisconsin pushing Obama over the top at 1:20 AM, as I predicted, it was Ohio at 11:13 PM.  NBC was more aggressive than CNN as best as I could tell, but everyone, including Fox, called it around 11:15 PM.  That is, everyone except Romney, who stubbornly held out for another hour with the hopes of pulling an “inside straight” with the first draw being a reversal of Ohio.  I've got some work to do on this one, but I believe I was the only one who attempted to "call the call"!

Of course I have a chart that summarizes all this.  Check it out, below.

Lots of you have asked what I plan to do with www.obameter2012.com after the election.  My current plan is to scale it way-way back, to maybe a post per month on an item of interest.  And then come back for the mid-terms in 2014, heading right on into the 2016 race thereafter!  In the meantime….maybe I have to think of a new name….

I’ll be back with a few more posts in the coming days before I put the election to bed…comments welcome!



Tom Gardner
Nate Silver
Actual
PRESIDENCY
Winner



States Called Correctly

49 out of 50 (FL)
49 out of 49

Swing states Called Correctly

7 out of 8 (FL)
7 out of 7

Electoral Vote
Obama
303 - 235
313 - 225
332 - 206 (FL to Ob)
Popular Vote %
Obama
 49.8 - 48.7
50.8 - 48.3
50.3 - 48.0
Popular Vote
Obama
62.7MM - 61.4MM
dnp
60.1MM - 57.4MM
Time race is called by NBC

1:20 AM EST
dnp
11:13 PM EST
State that pushes the winner over

Wisconsin
dnp
Ohio
Ohio
Obama
51 - 48
51 - 48
50 - 48
Nevada
Obama
51 - 47
52 - 47
52 - 46
Wisconsin
Obama
52 - 47
52 - 47
53 - 46
Iowa
Obama
50 - 48
51 - 48
52 - 46
Colorado
Obama
50 - 49
51 - 48
51 - 47
Virginia
Obama
51 - 48
51 - 49
51 - 48
New Hampshire
Obama
51 - 48
51 - 48
52 - 46
Florida
Romney
49 - 50
50 - 50
50 - 49 (likely Ob)
SENATE




Senate
Democrats
53 - 47
52.5 - 47.5
55 - 45
Races Called Correctly



31 out of 33
31 out of 33
Close Races Called Correctly



10 out of 12
10 out of 12
Missouri
McCaskill (Dem)
52 - 48
52 - 46
55 - 39
Ohio
S. Brown (Dem)
53 - 47
52 - 47
50 - 45
Connecticut
Murphy (Dem)
54 - 46
53 - 46
55 - 43
Pennsylvania
Casey (Dem)
53 - 47
53 - 45
54 - 45
Wisconsin
Baldwin (Dem)
51 - 49
51 - 49
51 - 46
Virginia
Kaine (Dem)
51 - 49
51 - 48
52 - 48
Massachusetts
Warren (Dem)
52 - 48
52 - 47
54 - 46
Indiana
Donnelley (Dem)
54 - 46
50 - 48
50 - 44
Arizona
Flake (Rep)
47 - 53
47 - 52
45 - 50
Montana
Rehberg (Rep)
49 - 51
48 - 50
49 - 45
North Dakota
Berg (Rep)
48 - 52
47 - 53
50 - 50
Nevada
Heller (Rep)
47 - 53
48 - 50
45 - 46
HOUSE




House
Republicans
196 - 239
dnp
194 - 233 (8 tbd)
Races Called Correctly


dnp
410 out of 427
Close Races Called Correctly


dnp
98 out of 108


3 comments:

  1. Impressive calls Tom . . . Nate better be looking over his shoulder in 2016 (with Hillary/Michelle on the D ticket as Jimmy Fallon implied?)

    ReplyDelete
  2. Very good predictions! You can help me in my applied regression and design class next semester! :-)

    ReplyDelete
  3. Your's program is extremely awesome and it is help people understand the significance of science, technology, engineering within their lives.This goal would be to prepare our youth for that jobs and skills for the future. I believe it is good simple for the type of problem-solving and inventive thinking necessary for daily existence.Discover! If you wish to find out more on the certain subject to spread your knowledges you should use custom writing.It may also assist you with your speech or presentation, research proposal, even creative writing.

    ReplyDelete

Leave a comment