This is my first DailyKos diary. I have lurked here for years. I have been flu-ridden for the past few days and have done some armchair quarterback punditry, which I thought I may as well post.
In any event, what I have done is try to predict the Pennsylvania margin by looking at the pollster.com polling averages in previous primaries as compared to actual results. I have looked only at head-to-head contested primaries (not caucuses, not when Edwards was in, not FL/MI), and only at those states where there was sufficient polling that I felt comfortable that the polling average was meaningful
Below is a chart comparing the pollster.com polling average to the actual results in each state that fit my criteria:
POLLSTER.COM ACTUAL DIFFERENCE
Georgia Obama +14 Obama +36 O+22
Alabama Clinton +1 Obama +14 O+15
Virginia Obama +18 Obama +29 O+11
Wisconsin Obama +7 Obama +17 O+10
Tennessee Clinton +22 Clinton +13 O+9
Mississippi Obama +16 Obama +24 O+8
Connecticut Clinton +1 Obama +4 O+5
Maryland Obama +21 Obama +23 O+2
Missouri Clinton +1 Obama +1 O+2
New York Clinton +19 Clinton +17 O+2
New Jersey Clinton +10 Obama +10 EVEN
Arizona Clinton +7 Clinton +9 C+2
Texas Clinton +2 Clinton +4 C+2
California Clinton +6 Clinton +9 C+3
Illinois Obama +36 Obama +32 C+4
Massachusetts Clinton +11 Clinton +15 C+4
Ohio Clinton +6 Clinton +10 C+4
As you can see, the pollsters on average have done remarkably well except in the South, where they vastly underestimated Obama over and over again, likely due to the impact of AA voters. They also blew Wisconsin, which I attribute to Obama riding high on a huge winning streak at that time and also to Clinton's half-hearted effort there, having chosen instead to focus on Texas and Ohio. Because those factors are not present here, I do not believe that there will be anything like the swing in Wisconsin.
What was most interesting was that Clinton never outperformed the polling average by more than four points. It was also interesting that Clinton tended to slightly outperform the polls in states where there are deep reservoirs of voters with Clinton-favorable demographics - a correlation that makes sense. I certainly believe that Pennsylvania is the kind of state in which Clinton has tended to slightly outperform the polls. It is far more comparable to the states where Clinton has outperformed than the states where Obama has outperformed. Therefore, I expect Clinton to outperform the pollster.com average by 0-4 points in Pennsylvania.
The current pollster.com polling average is 5.1 points. Based on that, I predict a Clinton victory of 5-9 points. If I had to guess, I would say Clinton wins by six. My prediction would change as the polling average changes. I do not see any basis in the history of this race for predicting a blowout Clinton victory, nor do I see a basis to believe that Obama will win unless the last-minute polling significantly changes.
For Clinton supporters, I recognize I have left out New Hampshire, where Clinton exceeded the polling average by 10 points. I also left out all caucuses, where Obama routinely exceeded his polling averages by enormous amounts. I am trying my best to compare apples to apples, and I do not believe that there is a good comparison to New Hampshire due to size, demographics, the presence of Edwards, and the remoteness in time. I do not expect a New Hampshire-size swing here any more than I expect a Wisconsin-size swing for Obama.
Anyway, I'm no Poblano, but welcome to what the world of what I do when I am home with the flu.