The short answer: the poll you like, compared to the aggregate.
The longer answer: Comparing one survey to another is like comparing apples to bowling balls – they’re super different because survey methodology differs from polling firm to polling firm...
one survey may be taken all online, while another is all via landline phones;
one may be talking to adults, while another is talking to likely voters;
one may be talking to a bunch more Democrats, the other a plurality Republican…
Further to the point, I’m often asked ‘which poll is most accurate?’ … Accurate how? At capturing Biden’s job approval? At being representative of the population as a whole? Of registered voters? Of likely voters? I find most people usually ask this question in the lead up to an election, and my response is always the same “the most accurate poll is the one that closest mirrors Election Day turnout.” … but, we generally don’t know that information until Election Day, so that answer doesn't play too well.
To be honest, I can make an argument for why each one of the below polls is the most accurate. I could also make an argument for why each one of the below polls is the least accurate.
So how should you find one you like? I’d start with the poll’s methodology and subquotas, meaning how the poll is conducted and who the poll is interviewing (the survey’s breakdown of age, race, ideology, etc.). Most polls these days will share:
how they are contacting folks (online, cell, landline);
what type of folks they are interviewing (adults, registered voters, likely voters); and
what percent of the people make up certain demographics (age, race, ideology).
Generally speaking, these are super different from polling firm to polling firm, but fairly stagnate among the same polling firm month to month.
If you look at a poll and don’t like that it’s done all online, or think it’s over-represented by Republicans, or younger folks, or left-footed Rastafarian unicyclists, then keep looking to find a poll which best matches your opinion of the nation. Then follow that poll month to month.
Once you have found a poll that asks the questions you like, amongst the folks you like, I suggest you find a polling aggregator (there are a million, but I typcally use RealClearPolitics myself), and see how the survey question or questions you follow match against the aggregate.
For example, if you favor The Hill/HarrisX poll in the above, great! But it’s important to know that while your favorite poll shows Biden’s approval is +22, this is a bit higher than the average. Is it wrong? Maybe not, but it’s important to know that’s higher than the average.
It’d also be great to know how it compares to the aggregate month in and month out:
Is it consistently 10-15 points above the aggregate?
Does it fluctuate in tandem with the aggregate?
Does this month’s aggregate have a big outlier in it that wasn’t there last month?
These are things that you’ll start to notice by following your one or two polls against ‘the average’.
So back to the short answer to ‘what political poll should I be following?’ I’d suggest following the one you are most comfortable with, and seeing how it stacks up against an aggregate of other polls, both in that moment and over time.
NOTE: If you want to get into the tons of other ways you can track a poll’s accuracy (like question order, margin of error, aggregate outliers, etc.), drop me a line and I’d be happy to chat about it!