538 says it has a R+0.8 House Effect
So is that (D internal) thing real or just editorializing?
FiveThirtyEight does label it a Democratic internal, so they assume by default it's probably biased toward the Democratic candidates. That's why their
house effect calculation moves it toward Cruz by 1.4%, although their poll rankings show that Change Research has a
mean-reverted bias of R+0.8.
They are very different things, as house effects judge polls against the average of the other polls (in other words, it tries to make all polls conform to the polling average, thus making the assumption that the average is right), while bias measures polls based on their performance relative to actual historical election results. I prefer to use the latter, so in my opinion this poll really is something to worry about for Cruz.
Of course, campaigns do have a tendency to release their very best internal numbers to try to shape public perception of the race, which is why publicly-released internal polls are generally viewed with suspicion.
