points, which is less than they were actually in 2012. in the states you see bigger polling errors in the midwest. you see clinton beat her polls in new york, in california and in oregon and washington. >> now that's interesting. i've been focusing on how state polls underestimated trump support by 7, 8%. that's way beyond the marginal of error. >> that means you rack up a huge number of popular votes and almost match obama's match popular vote so it's not simply a matter of the polls underestimating trump, not catching up to the demographic changes that are underway in the country. >> is there a bigger failure among journalists then, editors when we boil down the numbers into just a very simple portrayal? >> there are a couple things. one you can have a lot of polls say the same thing. at some point it doesn't tell you that much more. to have a hundred polls showing clinton two points ahead while