Last week I posted about a possible trend spotted in PA.
http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=132&topic_id=5260310The basis for this trend was Rasmussen, which was done on a near-weekly basis, probably done with the same methodology, and was showing a consistent trend of Obama closing the gap between himself and Hillary by 2 percentage points per week.
Well, this week he closed it by 5 in Rasmussen's poll, messing everything up.
Poll ................................. Date .............. # polled ...... Clinton ..... Obama ... Undecided .. Difference
PPP (D) ........................... 3/31-4/1/08 ..... 1224 LV ........... 43 .......... 45 .......... 13 .......... -2
Rasmussen ..................... 3/31/2008 ...... 730 LV ........... 47 .......... 42 .......... 11 .......... 5SurveyUSA ........................ 3/29-31/08 ...... LV ................. 53 .......... 41 ........... 2 ........... 12
Quinnipiac ......................... 3/24-31/08 ..... 1549 RV ........... 50 .......... 41 ........... 8 ........... 9
ARG ................................ 3/26-27/08 ........ 600 LV .......... 51 .......... 39 ............ 8 ........... 12
Rasmussen ..................... 3/24/2008 ...... 697 LV ........... 49 .......... 39 .......... 12 .......... 10PPP (D) ............................. 3/15-16/08 ..... 597 LV ........... 56 .......... 30 .......... 14 .......... 26
Franklin & Marshall College .... 3/11-16/08 ..... 294 LV ........... 51 .......... 35 .......... 13 .......... 16
Quinnipiac ........................ 3/10-16/08 ..... 1304 LV ........... 53 .......... 41 ........... 6 ........... 12
Rasmussen ..................... 3/12/2008 ...... 697 LV ........... 51 .......... 38 ........... 11 .......... 13SurveyUSA ......................... 3/8-10/08 ...... 608 LV ........... 55 .......... 36 ........... 3 ........... 19
Susquehanna ...................... 3/5-10/08 ...... 500 RV ........... 45 .......... 31 ........... 19 .......... 14
Strategic Vision (R) ............. 3/7-9/08 ......... LV ............... 56 .......... 38 ............ 6 ........... 18
ARG .................................. 3/7-8/08 ........ 600 LV .......... 52 .......... 41 ............ 6 ........... 11
Rasmussen ...................... 3/5/2008 ....... 690 LV .......... 52 .......... 37 ............ 11 .......... 15Darn that Obama anyway! Fortunately this latest poll doesn't take "BowlingGate" into account :D
So anyway, in the aforementioned thread I predicted a Clinton victory in PA:
Clinton 51.4%, Obama 46.6%, Other 2% - IF the Rasmussen trend held true. Well, it hasn't.
We have a couple of new wrinkles too.
I personally don't have a lot of confidence in the PPP polls - why? They just seem to always be so much out of the mainstream in poll results, and vary SO widely. I mean really - a 26-point lead for Clinton just over two weeks ago, and now a 2-point lead for Obama? So I'm sorry PPP, I'm excluding BOTH your results from my trend analysis.
Now we have two SurveyUSA polls taken about 3 weeks apart. Obama closed the gap 7 points in that time with SurveyUSA.
We have two Quinnipac polls taken about 2 weeks apart. Obama closed the gap 3 points in that time with Quinnipac.
And we have two ARG polls taken about 2 1/2 weeks apart. Obama lost ground by 1 point in that time with ARG.
It's hard to decide how to take this varying data.
Looking even further back on
pollster.com, We see that Quinnipac has been doing PA polls since early '07, as has Strategic Vision and Franklin & Marshall College. In reality, though, anything done before Iowa is pretty much ancient history.
For some reason Franklin & Marshall abandoned a 10-month-long history of only including Registered Voters in their polls, and switched to Likely Voters in their most recent one. Unfortunately, for me this is enough to exclude them from my trend analysis this time around.
Strategic Vision has stuck with Likely Voters. However, they've only done one poll since Iowa so they're also excluded. Emily's List is a single poll (commissioned by the Clinton-camp? Can't tell for sure) so I'm also excluding it. Muhlenberg, also a single poll and Susquehanna, only one poll since Iowa.
That leaves me with 5 Rasmussen polls, 2 SurveyUSA polls, 2 ARG polls, 4 Quinnipac polls.
Criminy. Quinnipac also switched last time, from Likely Voters to Registered Voters. What's wrong with these people?? Anyway, looking at Quinnipac's results, they're all over the place too (a 16-point gap on Feb 4, then a 6-point gap two weeks later, then back up to 12 points after another 2-1/2 weeks.) So to heck with 'em. (This is fun!) If they stick with Registered Voters next time maybe I'll include them again.
So I still have SurveyUSA, Gap closed by 7 points in 3 weeks, and ARG, gap widened by 1 point in 2-1/2 weeks. Dang it!
SurveyUSA is in line with Rasmussen, which has done some consistent polling lately and not changed its methodology. ARG is the only poll of them all that's shown an increase in the gap as opposed to a decrease so that's enough to exclude it too for purposes of trend analysis. (Really, this is a lot of fun!)
So we have SurveyUSA which seems to verify a 2-point closing of the gap per week, (but has only been done twice), and that darned Rasmussen poll which was going so well till Obama suddenly closed the gap by 5 points in the last week.
I'm just gonna split the difference (giving a little extra to Clinton, however, just to be conservative) and give Obama 3 points per week now.
We have 3 weeks to go since that last poll, so that would make the final total (averaging ALL polls INCLUDING the ones I discarded for purposes of figuring trend), AND giving 65% of the undecideds to Clinton just to be on the safe side:
Clinton 50.7%, Obama 47.3%, Other 2%So after all that, Clinton has lost 0.7% and Obama has gained 0.7% in the past week according to... um... ME!
If that translated into actual delegates, Clinton would win 82 vs 76 for Obama.
Somehow I get the feeling Hillary is hoping for a better result than this.