Wisconsin Governor: Crowley's Edge Narrows Depending on Sample

Marquette University's latest look at the Wisconsin governor's race has produced two results from the same fieldwork window, and they do not agree. One release, completed August 20, puts Democrat David Crowley ahead of Republican Tom Tiffany 49-44. A companion sample fielded over the same period shows the contest tied at 44-44, a result that holds across both the registered-voter and likely-voter screens. Same pollster, same days in the field, five-point swing in the topline.

The discrepancy is not necessarily a methodological failure — Marquette has a long history of releasing multiple cuts of its data, and house-level variance between samples drawn from the same population is a known feature of survey research, not automatically a red flag. But the size of the gap here is large enough that it complicates any straightforward reading of "momentum" in either direction. A pollster showing a candidate up five in one cut and dead even in another, within the same release cycle, is effectively telling readers that its own instrument cannot yet resolve the race to a level of confidence narrower than five points. That is itself the headline.

For Crowley, the two numbers matter differently depending on what campaign strategists are trying to signal. A 49-44 lead, if durable, would put him within range of a clean majority and would let allied groups argue that his standing has firmed up since the spring. A tied race, by contrast, undercuts any argument that he has built a floor above the mid-40s, and it hands Tiffany's campaign a data point it can use regardless of what other polling shows. Campaigns tend to publicize whichever number serves their narrative, and the existence of both within a single pollster's release gives each side a real, non-cherry-picked figure to point to.

The practical effect is that PollingSource's average continues to treat the race as a Toss-up, and this release does little to move it off that designation. Averages exist precisely to absorb this kind of intra-pollster noise rather than react to it, and with the tied sample sitting alongside the five-point Crowley lead, the net contribution to the average is closer to a wash than a shift. Readers looking for a clean signal from Marquette this cycle will need to wait for subsequent releases to see whether the tied result or the five-point lead looks more like an outlier once additional data comes in from other polling operations.

There is also a broader lesson here about how single releases get covered. A topline of "Crowley +5" is the kind of number that travels fast on social media and in campaign fundraising emails, but it tells only part of the story when the same organization's own data shows a tied race under a different specification. Until other pollsters weigh in with comparable fieldwork, the honest read on Wisconsin is that the race is competitive, the size of any Democratic edge is unresolved, and neither campaign has a clear data-backed claim to being ahead.