models are based on input. >> that's right. >> and in a sense it may be fair to say since the polls were wrong -- garbage in, garbage out. here's my question, though, this was not just a few polls. in the last five months i'm guessing there were about 150, 200 polls done. maybe trump was up in 15, 20 of those. hillary was up in the vast majority. vast, vast, vast majority. why were they so wrong? >> it's hard to say at this point but i think it's worth clarifying how they were wrong. it seems to me that the preponderance of the eviden preponderance of the polling error were found in the upper midwest of the united states. there were no polls that showed trump winning in wisconsin. the national polls don't look that bad, clinton will win the popular vote, the national poll said she was ahead by three or four. >> let's stop. a lot of people look at the "l.a. times" tracking poll, the