The ABC News/Washington Post poll is one of the very best public surveys available.

The specific numbers reported in its most recent iteration are probably not correct.

The broad conclusions it suggests are probably right on target.

Yes, all these seemingly conflicting statements are true at the same time.

A careful, thoughtful, and sophisticated methodologist, Gary Langer who conducts the survey for ABC and the Post is one of the best public pollsters laboring in this vineyard.

Nonetheless, the specific numbers in the most recent sampling, which generated a great deal of press attention, are probably not accurate.

Why?

One reason (among several) is that 100% of the interviews are conducted with respondents who answer their phones when they ring.

But these days, lots of people don’t answer their phones, especially their mobiles, when they don’t know who is calling.

Publicly available records of earlier ABC/Post polls suggest about half the cell numbers reached by callers resulted in “no answer,” compared to about ten percent of those dialed on landlines.

In our work, we’ve found that those who do answer their phones are different from those who don’t.

For example, phone answerers are more likely to be politically engaged while those who don’t answer are more likely to be Democrats and Biden supporters.

This may account for some of the anomalous results in the ABC/Post poll others have pointed out. For example, Natalie Jackson noted that Trump leading by 11 points among 18–39-year-olds is “nuts.” It is but can happen if you fail to reach young people who don’t answer their phones.

We don’t have large peer reviewed studies to prove this, or to validate the techniques our firm uses to circumvent the problem, so careful methodologist that he is, Langer sticks with what has been proven in decades past.

However, I’d argue that what doctors would call “our clinical experience” provides very strong support for the existence of the problem and the validity of the solution we employ.

Because the ABC/Post poll is less likely to capture Biden voters, I don’t believe the President is running 6-points behind Donald Trump or that his job approval is 37%. In fact, no other poll shows Biden that far behind Trump, nor an approval rating for the president that low.

Even including the ABC/Post poll, RealClearPolitics’ recent average suggests a tie, with Trump ahead by less than 1 point, and an average Biden approval of 42%

Three other recent polls put Biden in the lead by 2 or 3-points while another shows a tie. Only one other poll has Trump ahead.

Nobody would suggest any poll this month is somehow predictive of what will happen a year and a half from now.

But my broad takeaway from all these polls is the same—at this moment, the race for the White House is a jump ball. The national popular vote (which is, after all, the only thing national polls measure) could go either way.

That finding itself could shock many Democrats who, like me, believe President Biden is doing a great job and see Trump as that exquisitely rare combination of total evil and complete incompetence.

The simple fact is no poll suggests that judgment is shared by a majority of Americans. They all suggest a race that could go either way.

As we know, however, the national popular vote doesn’t decide presidential elections, the electoral college does.

Combining historical data with current polling puts President Biden ahead in states worth over 280 electoral votes—another clear indication he can win.

But several of those states sport razor thin margins, just like the national popular vote. While President Biden can win, there is certainly no guarantee he will.

That depends very much on what happens in the next year and a half.