With just over a month remaining before the 2026 midterm elections, a new examination of past Senate polling is drawing renewed attention to a familiar problem: how accurately do surveys measure the electorate before Election Day?
Patrick Ruffini of Echelon Insights recently examined Senate polling in Iowa, Michigan and Ohio at the same point in the election calendar—33 days before Election Day—in the 2018, 2020, 2022 and 2024 cycles. His analysis, published through The Intersection, compares polling averages with the eventual certified election results.
The historical numbers are notable.
According to Ruffini’s analysis, 93 percent of the polls examined overstated the Democratic candidate’s two-party support. The average error was 6.2 percentage points in the Democratic direction. Looking only at midterm elections, the average difference increased to 6.6 points.
