Oct 23 2016

10_23_16 General Election Polls: Toss Up, Trend Continues For Trump

Published by at 10:32 am under 2016 Elections

This is the 6th day I have looked at the general election polls for POTUS at RCP. For the 6th day the race looks to be a statistical tie or toss up. Today we have 4 polls , with one from ABC News that looks like an outlier (click to enlarge):

POTUS RCP 10_23_16

This is basically an uncomplicated assessment of the general election taken each day by averaging the % of the general election polls for each candidate. Here are the bottom line results for today:

  • HRC averaging 44.25%
  • Trump averaging 42.00%
  • Difference 2.25% (Clinton)

OK, we have a big change today (Clinton in the lead), but it is statistically still a toss up race.

Now let’s address that ABC News Poll. It is not only out of sync with the other three polls from today, it is out of sync with the other ELEVEN polls over the last 4 days (10/20-10/23). So it is questionable at best; and, therefore, so is the impact it has on the numbers today.

If we compute without ABC News we get:

  • HRC averaging 42.33%
  • Trump averaging 42.33%
  • Difference 0.0%

This is a big difference. This is also much more in line with what we have been seeing ever since the last debate.

So is it cherry-picking to remove an outlier? Not really, many of the more sophisticated poll-of-polls remove outliers , or at least lower their weighting (e.g Nate Silver’s 538). Note that all the other polls are daily tracking polls, and this ABC News poll probably comes out once a week at most. So it is a different animal at its core.

So let’s look at the trends, both with and without the ABC News “noise”. First without the outlier:

Gen_Elect_Trend_10_23

As we can see without the ABC News Poll very little has changed. There is a statistical tie with both candidates hovering around 42%. Which means there is 16% of the electorate hovering between undecided, a 3rd party candidate, not saying or not voting.  These are the voters who will decide the  election – and I doubt we will get an accurate read on them until the results come in.

Trump is still trending slightly up and Clinton slightly down, but the projections of the trend (dotted lines) are converging as each day we see about the same results. This shows the race is stabilizing at these levels and, bar something unforeseen, this is where it will be when the voting ends.

So let’s now add in the ABC News Poll:

Gen_Elect_Trend_10_23_v2

We see how one outlier poll can move a result quite a bit in a single day. But only for that day.  The trend lines demonstrate how the previous polls we have been sampling, when included, actually carry more weight than the one outlier.

So what does this mean? Same story since the debate. Basically tied, small trend for Trump.

I expect more polls may come out today – so if I get a chance I will update later this evening. If the ABC News poll is not an outlier we will see more general election polls moving towards HRC. I can use either path going forward (with or without ABC News).  We need to see a few days before we pick our path.

One final note. If you are wondering why my numbers are so different from the RCP numbers it is most likely due to the fact that I do not differentiate between the 2-way and 4-way polls. By including both types of polls I get a clear picture of the die-hard Trump and Clinton support, which is amazingly consistent. I believe this is a more solid number, because:

  • 3rd party candidates always fade out on election day as voters’ wishes must face reality. The more stark the choice, the more 3rd party candidates fade for serious (regular) voters. This is a very “stark choice” election.
  • Turnout models in all these polls are basically “experimental” this election cycle. With the Trump effect, the Bernie effect, the historic turnout in the GOP primaries and the major shifts in party registration, it is hard to tell who will come AND how they will vote. This has been a caveat from all the pollsters and poll-of-polls analysts. It adds yuge uncertainty to all polling results.

In a post later today I will show why I think the turnout models for Florida are incorrect based on the Vote-by-Mail statistics. (Update: here is the post)

My previous posts were:

One response so far

One Response to “10_23_16 General Election Polls: Toss Up, Trend Continues For Trump”