Final Week Polling Leads
Year: Polling Lead : End Result : Shift
2020: Biden+9.1 : 5.6 - 12.7 (95% percent confidence range) : ?
2016: Clinton+4.8 : Clinton +2.1 : -2.7
2012: Romney+0.5 : Obama +3.9 : -4.4
2008: Obama+7.0 : Obama +7.3 : +0.3
2004: Bush+2.2 : Bush +2.5 : +0.3
2000: Bush+3.4 : Gore + 0.5 : - 3.9
1996: Clinton+14.7 : Clinton +8.5 : -6.1
1992: Clinton+10.0 : Clinton +5.6 : -4.4
1988: Bush+11.8 : Bush +7.7 : -4.1
1984: Reagan+17.8 : Reagan +18.2 : +0.4
1980: Reagan+1.3 : Reagan +9.7 : +8.4
Big final week shifts are possible, but:
* they usually follow an established trend (2016, 2012, 1988, 1980)
* require a large number of undecideds or general voter ambivalence (2016, 2000, 1996, 1992)
* feature strong 3rd party candidates (2016, 1996, 1992, 1980)
* feature major events close to the election (2016, 1992, 1980) - 2008 wouldn't count here as the economy collapsed in September
* involve some potential polling error (2016, 2012) I say "potential" because there was a trendline toward the final result which the polls reflected accurately
I think Biden's final national lead will hold fairly steady. Worst case is around 5.5, which could put a Trump EC win in play if Biden collapses in one or more of the Rustbelt Trinity, and his best case is a little over 12.5.
