The most illuminating poll discussion I’ve seen about Alabama is a SurveyMonkey report showing that small differences in how they weighted their sample and modeled the electorate generated outcomes ranging from an 8 point win for Jones to a 9 point win for Moore. This is more the norm than you might imagine once you look under the hood of a lot of polling. But this is magnified greatly for this race because of several very hard to predict factors.
First, it’s a special election. Those are normally very low turnout and have outcomes that are hard to predict. Alabama has a Democratic voter base which is accustomed to having close to no chance in statewide elections. How does a possible win affect their motivation and propensity to turn out at the polls? Perhaps most important, how will all the news of the last month affect enthusiasm and turnout among would-be Moore backers? Sure, the diehards are on fire for Moore. But those don’t get you to a majority.
Finally, Alabama is not Virginia. Even Virginia isn’t really Virginia anymore. But in Virginia, we saw that even the Democratic campaigns seriously underestimated Democratic intensity and enthusiasm. These changes are quite difficult to model. Pollsters could be underestimating that factor too.
Of course, at the end of the day, Alabama is one of the most Republican states in the country. It’s a special election but it has also had a massive amount of attention not just in the state but around the country. Who wins could also determine the fate of the GOP tax bill. The specifics of the bill may be disastrous, even for many Trump backers. But the bill itself has become tightly associated with GOP party affiliation. That matters.
What this all comes down to is that I’m not sure these polls are really telling us much at this point. They are less different data points than various efforts to model the electorate and predict turnout, any of which could be right or wrong. There’s a huge possible range of outcomes tomorrow night and little way to know which is most likely.