Key Highlights

  • A candidate trailing by 22 points in polling won Wisconsin's Democratic gubernatorial primary by 0.5%, following a similarly large polling miss in Michigan's primary last week
  • Polling experts say primary elections are especially hard to forecast because they draw a small, highly engaged electorate that skews toward frequent poll respondents
  • Falling response rates have pushed pollsters toward online, opt-in methods, which tend to overrepresent politically engaged respondents and are vulnerable to bot activity
  • A similar dynamic contributed to a major polling miss in last year's New York City mayoral primary, where the eventual winner overcame roughly a 10-point deficit

Wisconsin Primary Upset Follows Similar Miss in Michigan

A candidate trailing by 22 points in pre-election polling won Wisconsin's Democratic gubernatorial primary by 0.5%, an outcome that has renewed scrutiny of polling methods less than three months before November's midterms. The result follows a comparable polling miss in Michigan's primary the previous week, prompting pollsters to warn that primary elections are proving especially difficult to forecast this cycle.

Polling experts say the challenge stems from the nature of primary electorates, which are typically small and composed of highly engaged voters. When one candidate draws disproportionate support from college-educated, politically motivated voters, the group most likely to respond to polls, it can produce a significant gap between survey results and the eventual outcome. Recent upsets may also have simpler explanations tied to individual candidates, including viral Social Media controversies that surfaced late in the race and may not have been captured in earlier surveys.

Response Rates and Turnout Models Under Strain

With randomized telephone response rates in the low single digits, pollsters have increasingly shifted to online, opt-in survey methods that are cheaper but attract a narrower, more politically engaged pool of respondents. Some of these respondents may not be human at all, given the growing presence of automated accounts on survey platforms, a concern several polling experts have raised as a Factor complicating accuracy.

Because primary elections lack the reliable weighting variable of party affiliation used in general elections, pollsters must estimate turnout using less consistent assumptions, increasing the risk of a significant miss. A similar dynamic contributed to a major forecasting error in last year's New York City mayoral primary, where the winning candidate overcame an approximate 10-point polling deficit, an outcome attributed to unprecedented youth turnout that existing likely-voter models failed to anticipate.

Pollsters have spent more than a decade adjusting for under-represented groups, a practice that intensified after 2016 revealed that less-educated Republican voters were less likely to respond to surveys. Weighting adjustments introduced since then have not fully resolved the underlying problem, since even reweighted samples can miss voters whose political preferences diverge sharply from those who do respond.

Implications for November's Midterms

Current polling shows Democrats favored to retake the US House of Representatives in November, but the recent primary misses have raised questions about whether similar forecasting problems could affect the general election. Experts caution that general elections benefit from larger, more representative turnout than primaries, which may limit how directly these lessons apply. Still, they note that despite ongoing accuracy challenges, polling remains one of the few tools available for gauging broad public sentiment ahead of major elections, even as confidence in its precision continues to erode.