Search This Blog

Sunday, August 13, 2017

The Polls Melt Down Again

A few months ago, the spread among the polls when it came to President Trump was -- to use a favorite word of the President's -- HUGE.  Some polls had a majority approving of Trump's performance while others had big majorities disapproving.  Then for the last month or two, most of that gap disappeared.  Well, it's back, and in a big way. 

In the latest Gallup poll, voters disapprove of Trump's job performance by 23 points.  In the latest Rasmussen poll, voters disapprove by a margin of 8 points.  That difference of 15 percentage points is much more than statistical error.  The margin of error in these polls is about +/-3.5%.  That accounts for only 7 of the 15 points at most (and likely much less).

So what is the difference?  Some, if not all of it is sampling error.  Rasmussen adjusts its results to get a sample group composed of the same mix of voters who actually cast a vote last November.  To be clear, Rasmussen is guided by the self-identification of the voting public in 2016.  Just because a Republican voted for Clinton, he or she is still counted as a Republican.  Similarly, if a person calls themselves a Democrat but then voted for Trump, that person is counted as a Democrat.  Gallup, on the other hand, does not adjust its numbers.  That means that if it happens to oversample Democrats, there is nothing done to correct that mistake.  Similarly, if Gallup gets responses from a disproportionate number of Republicans, no correction is made.  These different methods used by the pollsters may account for much of the difference in the numbers.

Given the different methodology used, which of the two polls is more accurate?  That boils down to a simple question:  has party identification changed much since November of 2016?  In other words, has there been a tidal movement from one party to another?  If not, then the Rasmussen method seems more accurate.  If so, however, then Gallup may have picked up a swing that Rasmussen missed.

How can we judge changes in party identification?  One way is to look at actual voter registration.  Across the USA, there has been a relative increase in the number of registered Republicans in 2017 compared to Democrats.  There may be a lot of publicity from the #Resistance, but the voters actually registering or changing their party ID have been moving towards the GOP not the Dems. 

Given that the only concrete measure of party ID has gone in favor of the GOP, it seems highly unlikely that the Gallup poll numbers are accurate. 

No comments: