looks like the polling error was bigger than 1948
Not sure about that. The problem with 1948 is that they simply stopped polling weeks before the election. So it's not really a fair comparison to now. Crossley and Gallup had Dewey up 5 and 6 points in their last polls. The actual result was Truman +5, so that's a net error of 10-11.
Roper had Dewey +15, but since we like to toss out outliers...
This election won't even be a larger polling error than 1980. Although 1980 polling got the winner right, the margin was off 6-7 points. The 2016 polling error looks like it's going to land in the 3-5 range right now.
