Phone polls undersample young people (18-35 - esp those who don't live with older family members), nonwhite people, and men. The first two of those groups are quite pro-Kerry; the last, pro-Bush. You can partly correct for the undersampling with statistical technique, but obviously the bigger the adjustment, the greater the likelihood of error. So polls could be significantly understaing Kerry's strength.
Another problem: likely voter models. Pollsters use a variety of factors to predict the likelihood of a respondent actually casting a ballot - like past voting behavior and demographics. Then they rank all the respondents according to their probability of voting, and draw a line between the likely and unlikely based on some projection of turnout. With this election, there could be two problems with this technique - the historically nonvoting demographics may vote in larger numbers than the past (esp the young and nonwhite), and the turnout could be higher than expected (i.e., the likely voting models could be excluding people who actually will vote, esp the young and nonwhite).
Maybe not. But polling is looking unusually treacherous this year.
Doug