So, like most polls, their summer numbers were -6% from the Republican’s final number.
Perhaps it results from the effect of employed people taking vacations. It sure is curious.
Anyway I take it as an historical fact when looking at the current polling and adjust it accordingly for predictive purposes.
It might be instructive to compare the Pew Poll with a poll that conservatives find to be more balanced in terms of sampling equal numbers of Republicans and Democrats and only uses Likely Voter samples: Rasmussen Reports:
5/14-5/16 1500 Likely Voters +/-3.0 MOE 46 Romney / 45 Obama
6/4-6/6 1500 Likely Voters +/- 3.0 MOE 46 Romney / 46 Obama
6/18-6/20 1500 Likely Voters +/- 3.0 MOE 47 Romney / 43 Obama
7/2-7/6 1500 Likely Voters +/-3.0 MOE 45 Romney / 45 Obama
7/12-7/14 1500 Likely Voters +/-3.0 MOE 45 Romney / 45 Obama
7/29-7/31 1500 Likely Voters +/-3.0 MOE 47 Romney / 44 Obama
8/8-8/10 1500 LV +/-3 MOE 46 Romney / 44 Obama
8/23-8/25 1500 LV +/-3 MOE 47 Obama / 45 Romney
9/ 7-9/9 1500 LV +/-3 MOE 50 Obama / 45 Romney
9/14-9/16 1500 LV +/- 3 MOE 47 Romney / 45 Obama
10/8-10/9 1000 LV +/- 3 MOE 43 Obama / 41 Romney
10/1-10/3 1500 LV +/-3 MOE 49 Obama / 47 Romney
10/4-10/6 1500 LV +/-3 MOE 49 Romney / 47 Obama
11/3-11/5 1500 LV +/- 3 MOE 48 Romney / 47 Obama