Polls must employ some kind of turnout model. One can call it Skewing if one thinks the turnout model overly optimistic for one of the choices. On election day we will find out which models are skewed...but msnbc is smoking hopium when they use a 2012 model.
Ah, but the 2012 election was the first to be “won” with the help of Diebold. As the country has done NOTHING to address the fraud “programmed” into the election the best model will be the one that accurately assesses ALL the important factors that impact the outcome. So the model that incorporates the influence of massive fraud programmed into these machines will best predict the outcome. It is sort of like gummint employment and other economic indicators. Who are you going to believe? Gummint or your lying eyes?