Q&A of the Day – Why does national polling look worse for President Trump than battlegrounds?
Each day I’ll feature a listener question that’s been submitted by one of these methods.
Email: brianmudd@iheartmedia.com
Twitter: @brianmuddradio
Facebook: Brian Mudd https://www.facebook.com/brian.mudd1
Today’s entry: Do you think it’s odd that most general election polls look worse for Trump than the battleground polling? Are there any pollsters you trust?
Bottom Line: I’ll start with the question most easy answered. I trust any pollster who gets sampling right. It’s the getting it right that’s the devil in the details. The single biggest issue with polling in this cycle has been the same as it was four years. Generally poor sampling. According to Gallup’s monthly party ID tracking, 1% more people ID as Democrats than Republicans currently. In Florida, based on voter registration, we know that Democrats outnumber Republicans by 1.3%. What this illustrates, is that any national sampling which samples greater than 1% more Democrats than Republicans isn’t a sound sample, and generously, no more than 2% more Democrats in Florida. So what do samples most recently look like?
Some pollsters don’t even disclose sampling, which instantly ensures a lack of credibility and many more bury it within their reporting. Let’s start with the NBC News/WSJ polling released Sunday (reported breathlessly by thrilled news media) demonstrating Biden with a 14-point lead. The first issue is with this poll is that they sampled registered voters as opposed to likely voters. That instantly presents a potential point of failure in these results. At this stage of cycle only likely voter samples should be used because it accounts for voter enthusiasm – which sampling registered voters does not. Second, the sample itself. Buried on the 12th page of the survey was the ID of those sampled. The sample was 45% Democrats compared to 36% Republicans. So here we have a poll which surveyed registered voters when it should have been likely, and they oversampled Democrats by 8% over national party identification. Still worse, the single largest block of individuals they spoke to were “Strong Democrats”. Hilariously this clearly flawed poll was assigned a 3.5% margin of error. You could quite literally double that margin and it still wouldn’t do the trick.
It’s more important that you don’t get hung up on a pollster, though there are some who are more responsible consistently than others, and instead read through methodology if you’re inclined. Accounting for samples compared to the current composition of voters is hard - that’s why I do what I do with my election related series (accounting for these factors). Now, regarding swing state polling vs. General election polling...what you’re describing is different than four years ago.
In 2016, Donald Trump polled an average of 4.4% behind Hillary Clinton in swing states on this date compared to 3.7% nationally. As we know that’s the inverse of what played out with Trump winning most swing states and Hillary Clinton winning the national popular vote. There’s a very interesting dynamic playing out currently. President Trump is polling several percentage points behind where he was nationally at this time four years ago, in part driven by the absurdity of polls like the NBC/WSJ cited earlier, however he’s actually polling around 1% better in swing states than four years ago. That’s an interesting tell. History is often a guide and given that this is an election won in the Electoral College rather than the national popular vote – President Trump appears to be slightly outperforming his 2016 path currently.