Primary polling is once again, usually sparse, unreliable and erratic and last night, it was no exception. We called the winner on every race, except the Arkansas Senate Runoff, where all polls showed a Halter victory but Lincoln ultimately picked up the undecideds. All other polling accurately placed the winner but missed the margins.
This post-game analysis will give an insight on the polling averages vs. the final outcomes and the best pollster of the race.
AR-Sen Runoff (Dem):
Metric: Halter 48 - Lincoln 45 (Halter +3)
Final: Lincoln 52 - Halter 48 (Lincoln +4)
Our metric was off by 7 points. There were only three runoff polls and both Kos polls showed an artificial momentum for Halter. The other poll done by Democracy For America was the most accurate, but that isn't saying much as it also predicted Halter to have the edge. Ultimately, incumbent Lincoln was able to pick up more undecideds than Halter to win the nomination.
Most Accurate Pollster: Democracy For America - Halter 48 - Lincoln 46 (Halter +2)
Metric: Whitman 57 - Poizner 26 (Whitman +31)
Final: Whitman 64 - Poizner 27 (Whitman +37)
Our metric was off by 6 points. Meg Whitman picked up the undecideds, not in as big of a way as the other races, but it was clear she had frontrunner momentum. Credit goes to Survey USA for being the first to pick up on the landslide.
Most Accurate Pollster: Magellan - Whitman 64 - Poizner 22 (Whitman +42)
Metric: Fiorina 44 - Campbell 22 (Fiorina +22)
Final: Fiorina 56 - 22 (Fiorina +34)
Our metric was off by 12 points. Carly Fiorina picked up the undecideds in a big way. Once again, credit goes to Survey USA for being the first to show Fiorina with a lead.
Most Accurate Pollster: Magellan - Fiorina 54 - Campbell 19 (Fiorina +35)
Metric: Branstad 52 - Van Der Plaats 29 (Branstad +23)
Final: Branstad 50 - Van Der Plaats 41 (Branstad +9)
Our metric was off by 14 points. This is not surprising due to the sparse polling in this race. Only two pollsters took a shot at this race: Selzer and PPP. Both showed Branstad with comfortable, but varying leads. Surprisingly, Van Der Plaats outperformed his polling numbers but ultimately fell short.
Most Accurate Pollster: PPP (D) - Branstad 46 - Van Der Plaats 31 (Branstad +15)
Metric: Sandoval 46 - Gibbons 29 (Sandoval +17)
Final: Sandoval 56 - Gibbons 27 (Sandoval +29)
Our metric was off by 12 points. The polling understated Sandoval's margin of victory. Gibbons performed close to our metric's prediction of his support, but Sandoval picked up undecideds in a big way. Another incumbent ousted in a primary.
Most Accurate Pollster: Research 2000 - Sandoval 48 - Gibbons 27 (Sandoval +21)
Metric: Angle 29 - Lowden 27 (Angle +2)
Final: Angle 40 - Lowden 26 (Angle +14)
Our metric was off by 12 points. Once again, Lowden's support was pretty much nailed in our metric but Angle picked up undecideds in a big way.
Most Accurate Pollster: Research 2000 - Angle 34 - Lowden 25 (Angle +9)
Metric: Haley 41 - Barrett 22 (Haley +19)
Final: Haley 49 - Barrett 22 (Haley +27)
Our metric was off by 8 points. Like in the previous two races, Haley, the frontrunner picked up the undecideds in a big way, missing a runoff by only one percentage point.
Most Accurate Pollster: PPP (D) - Haley 43 - Barrett 23 (Haley +20)