All pollsters normalize their data against a model of what they think the electorate will look like - number of R/D/I/O and a bunch of other factors.
Why do they do this? because they can't control the exact distribution of those who choose to respond plus people who are politically engaged are far more likely to both to spend time responding to a poll. Lots of little factors why they can't just use the data collected as-is.
This is where the rub comes in - republican leaning pollsters have data models that are more bullish on various R favoring factors, which is why they consistently slant R versus the rest. Here's the rub - they could be right. Trafalgar was the most accurate pollster in 2016, and in recent years was a A or B ranked pollster by 538 and they are hands down the most R biased. Why? Because we've had several elections over the past decade that have had big polling misses that understated R support.
Modeling the electorate is as much an art as it is a science, and unexpected aggregate misses do happen.
I do not agree with the approach of ignoring polls from right leaning pollsters, it gives a false sense of security. It's like claiming your football team was undefeated - if you ignore the games they lost.
= new reply since forum marked as read