CERC’s Polling Accuracy: 2018 California Elections

This analysis is based on the results from our 2018 California election polls. The primary election poll was conducted May 29 and 30 with a random sample of 504 likely California voters and a margin of sampling error of +/-4.4% at the 95% confidence level. The turnout model reflected 30% participation in that race. The general election poll was fielded October 29 through November 1 with a random sample of 428 likely California voters and a margin of sampling error of +/-4.7% at the 95% confidence level. The turnout model reflected 53% participation in that election. Note that we have not adjusted the turnout models to reflect actual turnout. All interviews were conducted by CERC’s trained professionals from our El Paso, TX, offices and via email.


Competitive Edge’s pre-election gubernatorial polls were highly accurate, predicting the exact percentages of the vote received by Democrat Gavin Newsom and Republican John Cox in both the primary and general election races. CERC’s results also correctly predicted the winner for U.S. Senate, as well as the outcomes for Propositions 6 and 10. These results are attributable to CERC’s innovative and rigorous methods, personnel training and deep commitment to providing accurate polling data.

Primary Election Results

Under California’s jungle primary system, the top-two vote getters—regardless of political party—move on to the general election. This could mean two members of the same party advance to November. Even in the difficult environment of a multi-candidate contest with six viable candidates and more than 30 certified candidates, CERC’s primary gubernatorial poll was highly accurate. CERC estimated the vote to the exact percentage point (undecideds accounted for) for the top-two vote getters who moved on to November, Democrat Lieutenant Governor Gavin Newsom and Republican businessman John Cox.

General Election Results

CA Governor
Our general election poll correctly predicted Democrat Lieutenant Governor Gavin Newsom would win with 62% of the vote and Republican businessman John Cox would finish with 38%. As the chart shows, our estimates accounting for undecided voters were spot-on and matched the actual statewide vote to the exact percentage point.

U.S. Senate
CERC correctly showed longtime incumbent Senator Dianne Feinstein easily beating California State Senator Kevin DeLeon, estimating Feinstein would win with 65% of the vote to DeLeon’s 35%. In the end, the margin was closer, 54% to 46%. Our overestimation of the gap in this race is due to two factors. A substantial 15% of the electorate was undecided, creating more play in the vote. Also, both candidates are Democrats which added a layer of complexity and led to Republican voters disproportionately supporting DeLeon (paradoxically the more liberal candidate).

Proposition 6
CERC predicted the hotly contested Proposition 6 outcome with a high degree of accuracy. Prop 6 would have repealed California’s 2017 Road Repair and Accountability Act, which had increased gas taxes and vehicle registration fees. The official ballot wording read, in part, “eliminates certain road repair and transportation funding.” Supporters criticized the official ballot language as misleading and campaigned to “repeal the gas tax.” We split our sample to test alternative wordings. Here we report the result for the version that used the actual ballot language.

CERC’s results projected 61% would vote against Proposition 6. Even with the much smaller sample size, our results were on target, as Prop 6 failed with 57% voting “no.”

Proposition 10
CERC correctly showed Proposition 10, which would have expanded local governments’ authority to enact rent control on residential property, would fail. With a large 19% undecided, this was another case where, with only a week left in the election, a lot of play remained in the electorate.


CERC conducted public polling on three important California races and two high profile statewide propositions during 2018. CERC’s results correctly anticipated the outcomes in all cases. Clients hiring CERC to conduct their private polling benefit from our proprietary sampling, weighting and data collection methods that lead to such a high degree of accuracy. At the same time, CERC avoids the mistakes other polling firms make, like quota sampling. We are currently evaluating engagements for 2020 campaigns. If you have such a need, please contact us at <> or 800-576-CERC.

pdficonCERC 2018 Polling Accuracy Memo