Debrief · 2018 cycle
The 2018 Debrief · post-mortem · permanent archive

How did the polling industry do this cycle?

The 2018 industry MAE landed at 4.5pt — graded across 480 historical races with a known outcome — and 373 polls were analyzed across 96 pollsters. Signed bias −0.42pt: polls landed close to the mean — no systematic over-predict on either side.

INDUSTRY MAE
4.5pt
vs midterm avg
SIGNED BIAS
−0.42pt
D-leaning
POLLS · POLLSTERS
373
96 firms
RACES SCORED
480
w/ known outcome
CYCLE 2012 2014 2016 2018 2020 2022 2024
Industry mean absolute error · 2012–2018
2012 4.7pt 2014 4.8pt 2016 6.5pt 2018 4.5pt 2020 5.9pt 2022 4.6pt 2024 4.1pt MAE · 2-WAY MARGIN
Per-pollster scorecard · 2018 · sorted by MAE
Signed bias · ±4.6pt
MAE · 0–5.5pt
st-pete-polls
−1.47pt
2.3pt
yougov
−0.53pt
2.5pt
cygnal
−1.92pt
2.5pt
mitchell-research
+1.03pt
2.7pt
suffolk-university
+1.04pt
2.8pt
marist-college
−1.02pt
3.3pt
emerson-college
±0
3.4pt
research-co
+0.69pt
3.9pt
missouri-scoutremington
−4.14pt
4.1pt
gravis-marketing
−2.67pt
4.2pt
quinnipiac-university
−0.95pt
4.2pt
the-trafalgar-group
+1.72pt
4.3pt
vox-populi-polling
−3.56pt
4.5pt
nyt-upshotsiena-college
−1.72pt
4.6pt
harrisx
+0.23pt
4.8pt
ipsos
+1.81pt
5.0pt
Biggest-miss races · 2018
Race Expected Actual Miss Direction
us-senate-2018-nm R+14.00 R+23.56 9.6pt D-margin underestimate
us-governor-2018-ca R+15.00 R+23.90 8.9pt D-margin underestimate
us-governor-2018-tn D+12.50 D+21.01 -8.5pt R-margin underestimate
us-senate-2018-ny R+25.00 R+33.44 8.4pt D-margin underestimate
us-governor-2018-ok D+4.00 D+12.10 -8.1pt R-margin underestimate
us-governor-2018-ny R+17.00 R+24.53 7.5pt D-margin underestimate
us-governor-2018-oh R+3.00 D+3.73 -6.7pt D-overestimate
us-senate-2018-me D+25.50 D+19.08 6.4pt R-margin underestimate