American Research Group, Inc.

 

2021 California Gubernatorial Recall Election Pollster Ratings

 

The following table lists the accuracy of pollsters in the September 14, 2021 California gubernatorial recall election using the measure of polling accuracy proposed by Martin, Traugott, and Kennedy.

An accuracy measure of 0.00 indicates that the odds ratio of the survey results matches the odds ratio of the actual results (the closer to 0.00, the better).

 

2021 California Gubernatorial Recall
Yes No Odds
Ratio
Accuracy
Berkeley IGS/LA Times 39% 60% 1.06 0.05*
Emerson 40% 60% 1.08 0.08 
PPIC 39% 58% 1.09 0.09 
Suffolk 41% 58% 1.15 0.14 
Change Research 42% 57% 1.20 0.18*
SurveyMonkey 41% 55% 1.21 0.19*
Stanford/YouGov 43% 57% 1.23 0.20*
Data For Progress 43% 57% 1.23 0.20*
KABC/SurveyUSA 41% 54% 1.23 0.21*
Targoz 42% 52% 1.31 0.27*
Redfield and Wilton 43% 51% 1.37 0.31*
Trafalgar 45% 53% 1.38 0.32*
Gravis 44% 49% 1.46 0.38*
CBS News/YouGov 48% 52% 1.50 0.41*
Actual 38.1% 61.9%

 

*The accuracy measures for 11 polls fall outside of their respective 95% confidence intervals.

The average absolute accuracy for all 14 polls is 0.217.

The average absolute accuracy for the 3 polls with accuracy measures within their respective 95% confidence intervals is 0.102.

The Berkeley IGS/LA Times poll had a sample size of 6,550 and therefore has a very narrow confidence interval.

While all 14 polls had the no vote winning, the average for the 14 polls missed the actual no vote by about 7 percentage points and the average for the yes vote missed the actual yes vote by about 4 percentage points. 

As a reference:

Joe Biden received 63.5% of the vote in California in 2020 and Donald Trump received 34.3% of the vote in California in 2020.

The average absolute accuracy for three California polls in the 2020 presidential election was 0.074. The Berkeley IGS 2020 poll had an accuracy of 0.19, which fell outside of its 95% confidence interval. The KGTV/SurveyUSA 2020 poll had an accuracy of -0.01. And the PPIC 2020 poll had an accuracy of -0.03.

The average absolute accuracy for all national polls in the 2020 presidential election was 0.085. The average absolute accuracy for all national polls in the 2016 presidential election was 0.070. The average absolute accuracy for all national polls in the 2012 presidential election was 0.055. The average absolute accuracy for all national polls in the 2008 presidential election was 0.037.

ARG Home