Here are the numbers to keep your eye on …
In yesterday’s post, we highlighted how sensitive “headline” poll results are to the assumed party-affiliation mix of voters.
Let’s add another piece to the puzzle: there are 2 sets of data that have to be closely watched when trying to make sense of the polls:
(1) the assumed party-affiliation mix of voters, and
(2) the survey-determined voter preferences by party affiliation.
I know that’s common sense … what I didn’t realize is how much those numbers vary from poll-to-poll.
To illustrate the point, let’s look at 2 polls: FoxNews (presumed to lean right) and Washington Post – ABC (presumed to lean left).
The most recent Fox poll had the race as essentially a dead heat … slight Clinton lead.
A recent WaPo poll had Clinton up by a couple of points.
OK,let’s play around a bit with the numbers …
What if we mix-and-match the Fox & WaPo numbers?
Let’s start by using the WaPo turnout assumptions (37%, 30%, 29%) … and the Fox party-affiliated voter preferences: Hillary gets 90% of the Dem vote, Trump gets 85% of the GOP vote.
Whoa, Nellie …
Suddenly, Clinton’s lead is up to almost 5 points.
Now, let’s flip the numbers around … use the Fox turnout assumptions (40%, 40%, 18%) … and the WaPo party-affiliated voter preferences: Hillary gets 86% of the Dem vote, Trump gets 88% of the GOP vote.
That puts a different paint job on it …
Trump up by almost 3 points.
What the teaching point?
I’m increasingly skeptical of the top-line numbers that are being reported.
Unfortunately none of the major polls give enough relevant detail to really decode what’s going on in their polls.
While I assume that the party-affiliated voter preferences are fairly legit, it’s increasingly apparent that the turnout assumptions are:
(a) widely variable … which makes sense since nobody really knows what the turnout mix will be, and
(b) the turnout assumptions – which are often buried in footnotes or kept secret — are carefully “managed” by the pollsters to get a result that’s comforting (to them)
Said differently, don’t take the polls too literally … they may mislead … either unintentionally or intentionally.