Nate Silver has 7 in 8 chance Dems take the house

[Awsi Dooger]

16. It will be interesting to see how close 538 comes to the House generic margin

Currently Nate Silver has it at 8.5% gap -- 50.6 to 42.1 -- while Real Clear politics is at Democrats +7.3.

The difference is adjustments. Nate amends polls based on the tendencies of the firm while RCP throws them up at face value. It can be tricky when dealing with adjustments. For example, 538 adjusts the Rasmussen generic ballot margin 3 points toward Democrats while the Trump approval adjustment with Rasmussen is 5 or 6 points lower. I think Nate would concede his generic model is in its infancy, and likewise the House and senate and governorship models. The models he is using now will be tinkered with prior to 2020.

The House generic polling hasn't been particularly accurate. The average error on RCP beginning in 2002 is 2.5%. In 2010 RCP badly overstated the GOP margin, finalizing at +9.4 when it ended up +6.8. It's easy to see what happened: All the double digit polls that don't resemble the real world shoved the number up to 9.4 in the final weeks. There were contributions from Rasmussen and others in the 10-15 range. Imagine how bad it would have been at +9.4, when Republicans managed +63 seats even with +6.8?

At least we are polling the generic ballot frequently by multiple companies as opposed to relying on polling in individual House districts, which are sporadic and unreliable to say the least. Generic margin is easily the greatest indicator, especially if the polling improves to determine what will play out. The 1994 red wave was a surprise largely because generic polling wasn't emphasized and the handful of polls that were done did not catch the Republican margin. I remember some polls a week or two out insisting that Democrats led. Only Gallup disagreed but they had it closer to even and not the +7.1 margin that played out, equating to 54 seats lost.