On Wed, 5 Oct 2016 14:18:07 -0600, Allah's Cock Bar wrote:

On 10/5/2016 2:04 PM, Ed Huntress wrote:
On Wed, 5 Oct 2016 10:50:54 -0600, Allah's Cock Bar wrote:

On 10/4/2016 10:13 PM, Ed Huntress wrote:
Well-done polls give you a snapshot in time. That's what we're looking
at now. It won't tell you who will win, but it will tell you how the
voting public is reacting to events.


Rasmussen Reports Clinton 41, Trump 42, Johnson 8, Stein 2 Trump +1

General Election: Trump vs. Clinton LA Times/USC Tracking Clinton
43, Trump 47 Trump +4


LA Times tracking is the outlier -- roughly 6 points under the other
professional polls for Clinton. Here's the story on it. They have an
odd weighting factor for people who voted in the 2012 election. The
consensus is that the LAT/USC level is an outlier but the trend lines
are good:

"Election Update: Leave The LA Times Poll Alone!"

http://fivethirtyeight.com/features/...es-poll-alone/

As for Rasmussen, it's a C+ rated poll with a +2 Republican bias.
However, it tends to call elections correctly (79%) *in the last week*
before an election. It's another odd one, being conducted with
robocalls.

Your best bets are either the Real Clear Politics averages:

The polls are inadequate this election due to the fear instilled in
Trump voters by the insanely rabid Dems.

I saw on TV a couple of days ago that some polling organization ran
some tests to see if there was such a factor at work, and they
concluded there wasn't. In fact, to the extent there is a small
"embarrassment" factor, it seems to work against Hillary rather than
Trump.

Their bottom line was that there is no evidence that the Democrats
have instilled "fear" in Trump voters. g


It's the same fear that has made bumper stickers disappear btw.

They've been in decline for years. There was a book about them
(_Bumper Sticker Wisdom: America’s Pulpit Above the Tail Pipe_), and
ten years later, the same author wrote an article about how much they
had declined.

You'll notice that Trump doesn't even offer them among his
merchandise.


Do not trust the polls.

It depends on what you're trusting them for. When the values in a poll
are pretty evenly divided, like 44%/46%, polls are no good as
predictors. When the values split widely, like 75%/25%, they're very
useful.

The usefulness of presidential election polls in recent years applies
to their detection of trends, rather than to absolute values. That
LAT/USC poll you brought up, for example, has NO sampling error from
one day's tracking to the next, because they keep asking the same
question of the same 3,000 people. If the numbers change, it tells you
something important about a trend.

--
Ed Huntress