Because it is possible that in some cases (but not in others) that the block of people not answering their phone forms a uniform voting factor.
If it were just a small group of people, this wouldn’t matter. But its not a small group of people anymore.
It’s now like trying to predict the temperature a bathtub will have when it fills while being unaware that three of the buckets being poured in are full of ice.
Oh they are?
Yes, why didn’t you check?
They wouldn’t talk to me.
If the % of people not answering cell phones were correlated in general with the probability of answering political poll questions one way, the Yougov polls might be wrong, but they are one of the most accurate. I think the Ras polling is more likely to be wrong because they limit it to landlines.
Maybe there is a such a correlation as you theorize in Texas in 2018. We will check the results, Ras vs Yougov.