Historical California ballot measures reveal persistent polling gaps

In 1996, polls indicated that California voters would reject Proposition 218 by a wide margin, yet the measure passed with 56.55% of the vote. This discrepancy serves as a stark example of how public opinion surveys can misalign with final election outcomes in the state. The Howard Jarvis Taxpayers Association has reviewed historical data on several major ballot measures to illustrate these gaps.

The analysis focuses on fiscal and tax reform measures, including Proposition 13, Proposition 218, and Proposition 5. The data suggests that polls often undercount support for taxpayer interests. This pattern has appeared across multiple election cycles, despite the use of various polling firms and methodologies.

Proposition 13 in 1978 offers an early case study in shifting voter sentiment. When Howard Jarvis filed over 1.2 million signatures in early 1978, initial polls showed a divided electorate. Support stood at approximately 25%, opposition at 27%, and nearly half of all voters remained undecided. By spring, as homeowners began receiving property tax bills, support increased significantly. The final Field Poll taken before the June 6 primary projected a passage margin of 57% to 60%. The actual result was 64.8% in favor and 35.2% against.

The gap between prediction and reality widened with Proposition 218, sponsored by the Howard Jarvis Taxpayers Association in 1996. Leading up to the November 5 election, consistent polling showed the measure failing badly. The final Field Poll released right before the vote reported that only 36% of likely voters supported the measure. Despite these numbers, Proposition 218 defeated expectations and won with 56.55% support.

More recently, polling on Proposition 5 in the 2024 cycle showed a similar, though smaller, divergence. The measure sought to lower the vote threshold for some local bonds from two-thirds to 55%. A Public Policy Institute of California poll reported that likely voters were evenly split, with 49% saying they would vote yes and 50% saying they would vote no. The proposition ultimately failed by a margin of 55% to 45%, a difference of roughly 5 percentage points from the projected tie.

These historical instances occur against a backdrop of current skepticism regarding polling accuracy. The Wedgewood Polls scandal has amplified concerns that some firms fabricate data. Wedgewood was added to major polling aggregators including The New York Times, FiftyPlusOne, and VoteHub, despite having no history or formal credentials. Its inclusion skewed aggregated results by mixing fabricated data with legitimate surveys.

The controversy over Wedgewood contrasts with errors made by established firms. A 2024 Des Moines Register poll conducted by J. Ann Selzer showed Kamala Harris leading Donald Trump in Iowa by three points. After Trump won the state by 13.2 points, Selzer was forced into retirement. The incident demonstrated that even respected pollsters with proven track records can produce results that diverge sharply from actual election outcomes.

Jon Coupal, president of the Howard Jarvis Taxpayers Association, notes that polling remains useful for measuring voter sentiment and message testing. However, he observes that polling conservative causes in California presents specific challenges. The historical data on tax measures suggests that many voters may conceal their fiscal conservatism or choose not to respond to pollsters. This behavior may explain why surveys consistently undercount support for taxpayer-focused initiatives.

The review of these ballot measures indicates that discrepancies between polling and actual results tend to favor a higher level of support for tax reform than surveys predicted. While the reasons for these gaps may include methodological issues or voter behavior, the data shows a recurring theme in California elections involving fiscal policy. The only definitive measure of public opinion, according to Coupal, remains the count on election day.