PVI Numbers for the last 4 elections. States trend upward, backward, sideways, down ways, and many other hidden factors that make trends hard to predict, but these are the raw numbers.
Year PVI % Change from Previous Election
2000: +2.22D Base Year
2004: +2.66D +0.44%D
2008: +1.61D +1.05%R
2012: +0.95D +0.66%R
Line of Best Fit predictor shows it at about +0.5D for 2016. Again, trends are very hard to predict and can speed up, stop, or reverse. Trends on a graph tend to look like a stock market, but you can get a tiny idea by looking at the numbers. Hope this data helps.
I was gonna ask why these numbers were different than the ones I have, but then I realized it must be because my numbers ignore third party and other voters. The national-popular-vote-corrected margins I have are as follows...
1996: D+0.69
2000: D+3.66
2004: D+4.97
2008: D+3.06
2012: D+1.52
In any case, the result gives a similar picture to yours--that is, that there is no clear and stable trend. The trends since 1980 would suggest a D+1.65 election in 2016, and the trends since 2000 would give D+1.22. In either case, I'd expect PA to vote about 1-2 points to the left of the national average in 2016.
The big difference is that some people inadvertently double all PVIs. The way Cook calculates it, 51-49 in a 50-50 election is D/R+1, not +2, for instance.
