"I am not in any way backpedaling," Matthews said. "I get frustrated when people think a poll that finishes a week out from election day is an absolute prediction of Election day outcome and vote spread." "I don't think Donnelly will win by 11 points, and the poll doesn't attempt to say that, but to try to explain via Twitter is hopeless," she added. The most recent Mourdock internal, conducted over the last two evenings, showed him leading Donnelly by 2 percentage points, 46 percent to 44 percent. That is well within the poll's margin of error of plus-or-minus 4 percentage points. While comparing surveys across pollsters can be problematic, the results for each survey are outside the other's margin of error, meaning that one of the polls would reflect the other's results fewer than 1-in-20 times. Matthews pointed to some key differences between the two polls that could have led to the disparity in results: the inclusion of cell-phone-only respondents and the composition of the poll according to age. "I would say we have a different opinion on the age composition (they have 56% over age 55) and we included 27% cell phones which, I think, does yield results that would differ if you only called landlines or had a small number of cells," Matthews said.