I want to make a mathematical point.
I have written many times over that when it comes to the numbers, I am
brutally neutral, and I would suspect that most people would notice this by now.
Numbers can be and are sometimes wrong, but the aggregates are usually on target, even if the margins are sometimes whacky. But I go where the numbers take me and simply report them.
Here were the the end-aggregates, as provided by RCP, for the 10 marquee Senate races of 2014:
KY: McConnell (R-inc) +7.2
AR:
Cotton (R) +7.0
CO:
Gardner (R) +2.5
AK:
Sullivan (R) +2.4
GA: Perdue (R) +2.8
IA:
Ernst (R) +2.3
KS: Orman (I) +0.8
LA:
Landrieu (D-inc) +5.7
NC:
Hagan (D-inc) +1.2
NH:
Shaheen (D-inc) +0.8
Of the six races where the aggregate showed a Republican winning, all six GOP candidates won. Caveat: I have not seen the end call for Alaska yet, but I assume that Sullivan will prevail, bringing the GOP to 53 seats.
The one aggregate showing an Independent winning was wrong: the Republican won, and pretty big, too.
Of the three aggregates showing a Democrat winning, the Democrats won in 2 of 3. Hagan lost. Landrieu indeed ended ahead of Cassidy, but probably to little avail: she is likely to be defeated in the runoff. And of course, Shaheen won.
So,
8 of 10 aggregates were correct, and they were from RCP, to note.
That being said, we also see a pattern of how far off the aggregates were, but only in general terms, for now:
in 3 of the 4 hotly contested Southern states, the aggregates were the most off in terms of a D vs. R race:
Kentucky, Arkansas and Georgia. However, it should be noted that the Georgia race is now less than an 8 point spread for Perdue, it is likely to land between +7.2 and +7.5 when the final canvasses are in, which would be right around Mitt Romney's margin in 2012. Still, that would mean that the aggregate would be off by 5, which is, of course, still too much. The aggregate was also off by about 6 points to the Left in
Iowa, at least according to the preliminary statistics. That is too much. Note that three of those four states are on the Mason-Dixon line.
The aggregates in both Colorado and Alaska were however, within bounds.
Also in North Carolina, although the call was false. The AK aggregate was Sullivan +2.4 and right now, he is winning by +3.4. A variance of +/-1 is well within bounds. In Colorado, the aggregate showed Gardner +2.5, the current winning statistic is +3.8 (with 93.3% of precincts in). A variance of 1.3 points is also well within bounds. The surprise this time is that the polls were off to the LEFT, where in the past, the polls had been off to the RIGHT in CO, NM, AZ and NV. In North Carolina, the aggregate showed Hagan +1.2, currently, Tillis has won by +1.5. That is a 2.8 variance, a bit much for my taste, but within the MoE. However, I tend to criticize anything outside of +/-2.5 points. Technically, it's acceptable. For my personal taste, it is not.
The aggregate in NH was off to the RIGHT: it showed Shaheen +0.8, she has won with +3.6 (current statistics), a variance of 2.8, the same variance as in North Carolina, at least with current figures. NH was the only aggregate with a bias to the Right. All others (for the 10 marquee Senate races, I mean), had a mathematical bias to the LEFT.
When the time is right, I will analyse this to great detail, but I always wait until the final canvasses are in,
because those margins that we are seeing right now are just preliminary, they are bound to change some, with last-minute absentee ballots, ballot challenges, etc. The Obama margin in Virginia in 2012 grew from +2.2 on election night (around 4 am) to +3.9 when all was said and done.
Speaking of Virginia, it wasn't even on my radar screen. I bet it was on practically no ones' radar screens, but quite obviously ALL polling was way off. That will be one for the record books, to say the least.
To see exactly how in-depth I go when I compare end-polling, take a gander here:
Statistikhengst s ELECTORAL POLITICS - 2013 and beyond The moment of truth how did the pollsters do
Please note the creation date of that report: Feb 4, 2013, 3 full months after the 2012 election. Most end-canvasses are done by December 16th of an election year (three gold stars for the person who knows why....), but then there are challenges usually in most every state and so I wait another month and sure enough, I usually see margins changed again. In the summer of 2005, the Alabama Bush raw vote from the 2004 presidential race was adusted by -10,000, because of a clerical error that no one caught until the summer of 2005. Similarly, the Romney raw vote count in Minnesota was adjusted downward by 3,000 votes 4 months after the 2012 election, for the same reason: simple clerical error.
In order for an aggregate to be so wrong, it can only mean that some polls were closer to reality and some were just unbelievably off. Even Rasmussen was off to the left: in the KS gubernatorial, it's final poll showed Davis +7, and yet, Brownback won, apparently, by +2. So, when the time is ripe, I will be scathingly harsh with ALL pollsters who were really off and believe me, if a pattern is to be discovered, I will discover it.
It's just as bad when polls are systemically off to the LEFT as when they are systemically off to the RIGHT. Neither is acceptable.
Still, it was fun to watch this election unfold and I congratulate the GOP on it's impressive win.
-Stat
