Why You Can’t Trust the Polls
Right now Biden is trashing Trump in all the polls. Biden’s leading in the national polls. He’s leading in a lot of state by state polls, which are more significant. I’ll explain why in a minute. And if you just count all the electoral votes, based on that, you’d say “Biden’s going to win.”
But there’s a lot of nonsense in that sentiment. First off, national polls don’t matter. Forget national polls because we don’t have national elections. We have state by state elections. Montana has an election, New York has an election, Texas has an election and so forth. A candidate wins so-called electors in each state. Then those electors get together in December and elect a president.
So the popular vote doesn’t ultimately count. Winning the popular vote doesn’t matter if you don’t win the electors. In fact, in 2000 and 2016, in Bush versus Gore and Hillary versus Trump, the person who was elected president lost the popular vote.
Biden will get five million more popular votes in California, but you can’t win California twice. You can only win it once. Meaning you just need one more vote than the other guy to win California. Well, if you have five million more votes, 4.9999 million votes are basically wasted. Again, you can’t win the state twice.
You only have to win Michigan and Pennsylvania and Ohio and Florida and some other places by one vote. That’s where the rubber meets the road. Now Biden is ahead in those state polls. But when you deconstruct the polls, you find out that they’re built in a way to favor Democrats and hurt Republicans. How do they do that?
The first question to ask is, whom are you polling? Whom are you asking who they intend to vote for? And you can select all adults. Well, if you’re pulling all adults, your poll is garbage because many will not vote at all.
Or you can limit your poll to registered voters. That makes sense, because if you’re not registered, you can’t vote. But that’s also a problem because 50% of registered voters don’t vote. Then you have to ask, “Who are the likely voters?”
That’s a subset of registered voters who voted in the last few elections and are likely to vote again. That’s a much smaller group. Only if you’re polling registered likely voters will you get a legitimate polling group.
But you’ll find that a lot of the polls don’t do that. So throw away the polls that aren’t focusing on the right group. And even within that group of likely voters, you have to ask, “How many Republicans am I going to ask and how many Democrats am I going to ask?”
That’s a fair question. There are more registered Democrats than there are registered Republicans. If you’re trying to be accurate, you would have more Democrats in your sample.
The question is, how many more? The proper skew is maybe 54% Democrat, 46% Republican, but these pollsters are all liberal so they’re doing 58% Democrats, 42% Republicans. And even within that sub-group, they can include more African-Americans than whites and African-Americans vote 90% Democrat.
It’s like tipping the boat to one side. In other words, pollsters are stacking the deck against Trump. You need to know all this to deconstruct the polling data and reverse engineer it in effect, which I do. Then you’ll see that the races are actually very close despite all the headlines. But the headlines are designed to demoralize Republicans.
I’ve actually done polling for presidential candidates, so I know how it works. Let’s get a little deeper into the weeds to see how polls can be misleading…
For 70 years, pollsters have been asking two questions. The first question is who you’re voting for. In this case your choices would be Trump or Biden. But you must then subject the answer to all the other disclaimers I just went through. But then they ask another question: Regardless of who you’re voting for, who do you think is going to win?
There you can get a different answer. Because someone could say, “I’m voting for A, but I actually think that B is going to win.” So the first question is called the intention question, for whom do you intend to vote, Trump or Biden? The second question is called the expectation question, which is regardless of whom you’re voting for, who do you expect to win?
Well, it turns out that from a data perspective, the expectation question has more information in it. Why? Because if you ask the expectation question, every person answering the poll is including a larger network of maybe 20 to a 100 friends, neighbors, co-workers, family members, et cetera.
So now all of a sudden your polling sample expands. It’s a more accurate poll because there’s more data in it. Right now, Biden is winning the intention question, where people say, “I’m voting for Biden.”
But Trump is winning the expectation question. People in effect are saying, “I’m voting for Biden, but I think Trump’s going to win.”
Importantly, the expectation question has a 78% accuracy rate. In other words, it has a better history of getting it right. Meanwhile, the intention question only has a 22% accuracy rate.
And Trump’s leading 55% to 45% in the all-important expectation question.
So the polls, if you properly understand how to interpret them, say Trump has a slight edge. That’s why you have to take the headlines with a grain of salt. I believe the election is going to be close. But seen from the important polling data, Trump has an edge. Not a huge edge, but a slight edge. Also, polls work with a lag.
It could take a week or more to sift through data. So, if Trump’s getting good news, which is admittedly rare, it can take 7 to 10 days for that to show up in the polls.
If Trump has a good week people say, “Fine, but Biden’s still ahead in the polls.” But those polls reflect the situation a week to 10 days prior. You’d have to wait a week and see if those polls say confirm what they were saying earlier.
It’s important to understand how all this actually works, which is why you can’t just accept what the headlines are saying. Again, based on the deeper polling data, right now I’d give Trump a slight edge. But only slight. It’s going to be a close election.
And the predictive models I use constantly need to be updated. Too much can change, one way or the other. So please don’t put a stake in the ground based on what I’m saying today. I’m only telling you what would happen if the election were held today, but we got 80 days left.
But get ready for a wild ride.
for The Daily Reckoning