On Polls and Polling

From this post:

The Hill report posts a YouGov survey (here) which, in section 4 (page 7 of the pdf), has the results of this question:

How likely is it that you will vote in the 2024 presidential election this year?
Among registered voters and those who plan to register to vote before November

The responses of note are
Definitely will vote: 79%
Likely will vote: 12%

For the math inhibited, that’s a whopping 91% of the registered voters surveyed believe they are at least likely to vote. That seemed to be a lot, to me. So, I wondered, how many registered voters voted in 2020?

Well, it seems that is not an easy question to answer, with so many Biden leaning districts having so many more votes than registered voters (I kid, I kid).

No, the number of votes cast for President in 2020 is an official number (158,429,631 according to the FEC. In that same report:

U.S. Census Bureau Voting Age Population (Current Population Survey for November 2020): 252,274,000
Percentage of Voting Age Population casting a vote for President: 62.80%

So, 62.8% of the voting age population voted. But, how many of those were registered (or even were able to register)? That is tough to answer, because the state’s voting rolls are overpopulated. This is because people die. People move. Some people aren’t citizens. Etc…

So, you can’t go by the number of registered voters in each state.

Well, it seems the US Census does try to answer such questions. They have issued a report with this table (a download of a CSV spreadsheet).

It is fascinating. It breaks down voting by age, race, gender, and a lot of other categories. This was done by sampling, so there is some associated error.

For example, the total number “Reported Voted” is 154,628,000 (not 158,429,631 , which is the official number). But, per their same sampling, the estimated number of 18 and over citizens is 231,593,000. Of those 231 million, 168 million are estimated to be registered. So, we have 155 million voters with 168 million registered, or (wait for it) … 155/168 = 91.7%.

That’s right, in 2020, nearly 92% of the people who could have voted, did vote. I find that amazing. It also falls in line with what the YouGov survey found for the 2024 election.

I guess this is why the push to register as many as possible. For example, the Swiftie effort. Of course, the 8% of the registered who didn’t vote could change the outcome, so there is also a need to make sure all people registered actually vote, but the percentages favor registering the largest number of people who are likely to vote Democratic as possible.

This is why I keep harping on the need to register 18 year-olds when so many of them are concentrated on college campuses in a few weeks.

Not really. You need to adjust for the people who could have registered to vote, but didn’t.

This is a smaller number than the total number of people eligible to register but who didn’t register, because I’m sure some tried to register but were prevented, or did register but got thrown off the voting rolls. But there are certainly people who don’t try to register: some because they don’t want to bother, some because they believe the false claim that that’s how you get picked for jury duty and they don’t want to be called for a jury, and some because they’re not going to vote anyway so why register?

Talking about polls, Reuters has a new poll showing Harris now ahead by 4 points (margin of error 2 points), getting stronger among women and Hispanics. Just for balance.

But wouldn’t you expect a correlation between people likely to vote and people likely to answer a survey?

From a national standpoint, she’s done well and it seems very likely she will get a higher share of the national popular vote. Nate Silver’s tracking of polling seems to confirm her advantage here by about 3.8% (as of August 29th, I’ll link below and this gets updated daily). He’s not a pollster himself, but he has a model that forecasts the election. She’s in the period right now known as the Convention Bounce. Usually, after conventions, the polls move in favor of the most recently nominated for a few weeks.

The question will be the Electoral College, where Republicans currently have a built-in advantage. It’s why even with a lead in the national vote, Nate forecasts a roughly 50% probability that she wins the election.

This is surely part of it. Someone with no interest in voting in this election probably isn’t going to answer 50 questions poised by a stranger on the phone. Registered or not.