2016 Election Polls Part 2

In my earlier post, I went throught the steps to how Trump picked up the Electoral Votes needed to win the Election over where Romney stood in 2012.

The common refrain after the election was that the polls were all wrong.  But were they?

Let's look at the polls from RealClearPolitics.  On Election Day the final projections were 203 For Clinton/Kaine and 164 for Trump/Pence with 14 states as well as 1 of the Maine Congressional districts a toss up.  These total 171 Electoral Votes.

Solid states were 115 to 63 Clinton over Trump
Likely states were 92 to 53 Trump over Clinton
Leaning States were 35 to 9 Clinton over Trump

Of these states only Wisconsin was picked incorrectly.  During the campaign Clinton consistently polled at +5 or better with the final RCP Average coming in at 6.5.   Trump finished +0.7 which was outside the Margin of Error of  all of the final polls which ranged from 1.9-4.9 points.

Keep in mind the Margin of Error is the 95% certainty that the result is within + or -  the value and the MoE.

Clearly either sampling in Wisconsin was bad or a lot of people changed their minds in the final week.

Now let's look at the overall result or what is known as the Popular Vote.  RCP Polls showed Clinton anywhere between 6 points up and 2 points down (1 poll).  The average was 3.3 points, the final result was Clinton +2.1.  The result was within the MoE of all the polls except for IBD/TIPP Tracking (Clinton +1.1 to Trump +5.1) Gravis (Clinton +5 to +3) and Economist/YouGov (which shows Clinton +4 with no MoE)   It should be noted that Rasmussen polling showed Clinton +2 which was the closest raw score to the final result.

The national polls were pretty much right on.

As for the Toss Up states I find it interesting to look at the RCP "No Toss Up" Map  While many websites were predicting a 70%-80% or more probability of  a win for Clinton, this map clearly showed a close race and one or two states could swing the results.

With No Toss ups, RCP showed Clinton winning 272 to 266.  But here is what everyone in the press ignored:

On October 16th, three days before the last debate, Hillary Clinton had a 340 to 198 advantage in the Electoral College and a 5.5% advantage in the Popular Vote polls.  At that point in time she maintained all of the Obama 2012 States (332 Votes) plus North Carolina (+15) while trailing in only Iowa (-6) and the Maine 2nd CD (-1)

On October 17th RCP moved Ohio to Trump (-18)
On October 19th the day of the debate Arizona moved from Trump to Clinton (+11)

Nothing else changed until October 30th, 9 days before the election, Florida moved to Trump (-29)  leaving her with a 304-234 advantage.  On November 1st and 2nd, Clinton picked up the Maine 2 CD (+1) but Arizona (-11), North Carolina (-15) and Nevada (-6) moved to Trump giving her a 273-265 advantage.

In the final days New Hampshire and Florida flipped back and forth and the Maine 2 CD came back to Trump.

So on 11/8 RCP said 272-266 Clinton in the Electoral Vote.

Actual Election day results only changed the following (in order of the calls):
Nevada (+6) Trump to Clinton  278-260
Pennsylvania  (-20) Clinton to Trump 258-280
Wisconsin (-10) Clinton to Trump 248-290
Michigan (-16) Clinton to Trump 232-306


We discussed Wisconsin earlier, Nevada is the only state of the four that RCP moved from one side to the other flipping it three times after Labor Day.

Nevada's final RCP average was 0.8 for Trump.  Clinton won by 2.4 or a difference of 3.2.
Pennsylvania's final RCP average was 1.9 for Clinton.  Trump won by 0.7 or a difference of 2.6
Michigan's final RCP average was 3.4 for Clinton.  Trump won by 0.3 for a difference of 3.7.

The pollsters took a lot of flak for "missing" this election but they were right on for the most part.

The 2016 National Result was off 1.1 points from the RCP polls (3.2 Actual to 2.1 Predicted)
The 2012 National Result was off 3.2 points from the RCP polls (3.9 Actual to 0.7 Predicted)
The 2008 National Result was off 0.4 points from the Rcp polls  (7.2 Actual to 7.6 Predicted)
The 2004 National Result was off 0.9 points from the RCP Polls (2.4 Actual to 1.5 Predicted

Reality check is that the polls were fairly accurate in 3 of the last 4 elections and not way off in the 4th.  In 2016,  the interpretations of a runaway election for Clinton fit the mold of what many wanted to believe, rather than what was actually happening.

Pollsters for the most part do a fairly good job.  The media should stop blaming them for being wrong when they weren't.