Polling Data Often Undercounts Support for Taxpayer Measures in California

Voters in California may face a disconnect between what pre-election surveys suggest and what actually happens at the ballot box. For those tracking fiscal policy, this gap can alter campaign strategies and budget projections. Recent data indicates that polling errors are not random but tend to undercount support for measures that limit government spending.

Jon Coupal, president of the Howard Jarvis Taxpayers Association, points to a specific pattern in the state's voting history. He notes that discrepancies between poll numbers and final results consistently favor the anti-tax position, even when surveys show the opposite. This suggests that a portion of the electorate may be less responsive to standard polling methods when it comes to financial issues.

The most recent example involves Proposition 5, a legislative effort to lower the vote threshold for certain local bonds. The measure sought to change a constitutional requirement in place since 1879, reducing the necessary approval from two-thirds to 55%. Just before the election, a Public Policy Institute of California poll found the electorate evenly split. The survey reported 49 percent support and 50 percent opposition among likely voters.

The final tally did not match that near-even split. Proposition 5 failed with support hovering above 55 percent and opposition at 45 percent. This represented a polling error of approximately five percentage points. The result highlights how close the preliminary data was to the actual outcome, yet how wrong it remained regarding the final margin.

Coupal identifies Proposition 218 in 1996 as the starkest failure of prediction in this category. That measure, also sponsored by his organization, aimed to reform taxation. Leading up to the November 5, 1996, vote, polls consistently showed the measure losing badly. The final Field Poll released immediately before the election projected only 36 percent support from likely voters.

Despite the negative data, the measure passed with 56.55 percent of the vote. This created a swing of nearly 21 percentage points between the last published survey and the final count. The result stood as a significant deviation from the established trend of polling accuracy for that cycle.

The pattern dates back to 1978 with Proposition 13. Early polls showed voters deeply divided, with support around 25 percent and opposition at 27 percent. As homeowners received tax bills in the spring, support rose. Even the final Field Poll before the June 6 primary underestimated the margin, predicting passage by 57 to 60 percent. The actual result was 64.8 percent to 35.2 percent.

Coupal argues that these historical gaps suggest many voters conceal their fiscal conservatism or simply do not respond to polls. He states that the only poll that truly counts is the one conducted on election day. This perspective contrasts with the current national debate over polling reliability, where the fabricated data from Wedgewood Polls has cast a shadow over aggregation methods used by major media outlets.