Wednesday, November 5, 2014
How Bad Were the Polls?
First, the Naked Dollar's report card. Our calls were:
+9 in the Senate
Actual result: +9 (assuming Cassidy wins the LA runoff, which he will)
+11 in the House
Actual result: +14 with four more too close to call (all of which would be GOP pickups)
Actual result: +3
Regarding the governorships, holy cow. No one saw that coming, and in states like Massachusetts and Maryland, no less. And no one saw the GOP picking up 14 house seats (with the potential for 18). After netting 63 four years ago, their upside was limited.
A wave, indeed.
So, how bad were the polls? Pretty bad, as it turns out. I did a quick analysis, comparing the actual results with the RCP polling average in each race. The results are as follows...
The average poll in the Senate missed the mark by 5.5 points. Out of 36 races, the polls underestimated the GOP result in 24.
The average governor poll missed by 5.4 points. Out of 36 races, the polls underestimated the GOP results in 29.
Bottom line: the pollsters missed by a lot, and they systematically underestimated GOP turnout (or overestimated Democrat).
No, I don't think they were biased. Every election usually produces a bias, but there is no consistent direction. Poll is a science, but an inexact one. Heck, the polls for the Virginia Senate race were off by 10 points. The lesson here is that they are useful, but only to an extent.