10
Loading video...
By Sam Stevenson
Associate News Editor
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
Democrat Scott Colom pulled off a huge upset in the Mississippi Senate race after every legitimate public poll and well-known political prognosticator favored incumbent Republican Cindy Hyde-Smith, with most classifying the contest as noncompetitive.
The race was officially called for Colom on Wednesday, when he led Hyde-Smith by 4,669 votes out of 962,237 counted, or 49.2 percent to 48.7 percent.
Before the race was called, Colom had taken issue with well-known political forecaster The Cook Political Report and other “prognosticators,” saying that their ratings made it more difficult to get funding from national Democrats and noting that the South had “been neglected for so long.”
Read More on Politics
His victory in the deep red South raises two questions. How did analysts communicate uncertainty in such a lightly polled state? And did those ratings discourage investment in a campaign that clearly had a viable path to winning?
Before Election Day, Costas Panagopoulos, chair of political science at Northeastern University, told Newsweek that a survey showing one candidate ahead should not be read as a meaningful lead when the candidates’ statistical ranges overlap.
“A 48-to-44 race that’s within the margin of error is not necessarily implying a lead for one of the candidates,” Panagopoulos said. “Statistically, it’s a dead heat.”
His warning proved especially relevant in Mississippi.
Where Polls And Forecasts Went Wrong
- The race’s final legitimate poll showed Hyde-Smith ahead by 3 points, within the survey's 3.5-point margin of error.
- Cook Political Report had shifted Mississippi from Solid Republican to Likely Republican, but never placed it in a more competitive category.
- Nationally, polls understated Democratic support by an average of 3.4 points across 88 contests reviewed by pollster G. Elliott Morris.
- A fabricated poll showed a close contest in the Magnolia State, but ultimately was not a real survey and was presented by a company saying it was illustrating weaknesses in media and polling-industry vetting.
Colom Says Influenced Funding in the Race
In a November 9 social media post, Colom accused forecasters of being “only a reflection of the past” and said Solid Republican assessments made fundraising considerably harder for his ultimately viable campaign.
His complaint raises a structural issue.
Ratings often influence where donors, parties and outside groups direct resources; weak investment can then become evidence to others that an underfunded challenger lacks a credible path.
Colom’s argument does not prove that Cook Political Report’s rating cost his campaign money, but it reveals that a winning candidate thinks an ultimately incorrect descriptive label can influence decisions by donors and political groups.
Newsweek contacted Cook Political Report’s senior elections analyst, David Wasserman, for comment but did not receive a response before publication.
...
What The Mississippi Polls Showed
A Data for Progress survey conducted from August 5 to August 15 among 1,018 likely voters showed Hyde-Smith leading the race with 44 percent of the vote compared to Colom's 38 percent, with independent Ty Pinkins getting 7 percent and 11 percent saying they were undecided.
The poll used SMS, live-caller and web-panel respondents and had a margin of error of plus or minus 3 percentage points.
But that result also contained a warning for the Republican incumbent. Fifty percent said they viewed Hyde-Smith unfavorably, while 57 percent had not heard enough about Colom to assess him.
Those results revealed a race with a defined incumbent facing a challenger who still had room to introduce himself to voters.
The only public poll during the final three weeks of the race came from Change Research.
That poll did show a tight race that might not be safely Republican.
Conducted online from October 21 to October 22 among 842 Mississippi voters, the survey showed Hyde-Smith at 43 percent, Colom at 40 percent, Pinkins at 6 percent and 9 percent undecided. Again, Colom was within the plus-or-minus 3.5-point margin of error.
...
The Likely-Voter Problem
Pollsters do not merely record preferences. They must decide whose preferences will matter on Election Day.
“Deciphering who the likely voters are and being accurate about that for Election Day is as much an art as it is a science,” Panagopoulos told Newsweek, noting that firms use different screens to determine which respondents enter their reported samples.
That challenge becomes more consequential when turnout departs from recent patterns.
The American Association for Public Opinion Research, a professional organization for polling and survey research, says some respondents who claim they intend to vote ultimately stay home.
A study published in Public Opinion Quarterly found that likely-voter screens can introduce error because respondents do not accurately predict whether they will participate. In its analysis of 2016 survey data, only 64 percent of people who said they would definitely vote had a validated turnout record.
...
Why Forecasters Stayed Republican
Cook Political Report rated Mississippi Likely Republican in its final November 2 assessment, placing it outside the Lean and Toss Up categories.
It was part of a wider consensus. Sabato’s Crystal Ball and Inside Elections also ended up at Likely Republican after moving the seat from safer GOP classifications during October.
The quantitative forecasts expressed even greater confidence.
Decision Desk HQ gave Hyde-Smith a 91 percent chance of winning and projected a 14-point Republican margin. Its model drew 64 percent of its forecast from fundamentals, compared with 11 percent from polling and 25 percent from prediction markets.
Cook defines Likely races as not competitive at that point, although capable of becoming engaged.
Its state analysis acknowledged that Colom had run “an admirable race” and identified the possibility of surging Black turnout, while concluding that Republicans believed that a “late influx of cash has fixed their problems,” suggesting that the forecaster gave this consideration added weight in its decision.
It also noted that Hyde-Smith had doubled down on her widely mocked response to rising beef prices, telling consumers they had “so many proteins to choose from.”
Colom’s campaign turned that remark into a viral ad, viewed more than 20 million times at the time of writing, in which voters said they would rather “pick another senator,” making affordability a memorable closing argument against the incumbent.
National Polls Underestimated Democrats
G. Elliott Morris, co-founder of FiftyPlusOne, reviewed 373 polls covering 88 contests and found that they underestimated Democrats by an average of 3.4 points. The historical average since 1998 was about 2.8 points in either direction.
The error was surprising partly because recent cycles had conditioned analysts to expect Republicans to outperform surveys.
Morris argued in an opinion article for MS Now that forecasters “whiffed on Mississippi,” though pollsters’ overall performance remained within the range recorded in several previous elections.
He said recent Republican overperformances had hardened into a misleading assumption that surveys would again understate GOP support.
“Expect the polls to miss by a few points, and don’t assume you know the direction,” Morris wrote.
...
The Fake Poll That Looked Plausible
Mississippi’s thin information environment was further complicated by Wedgewood Polls, which published fabricated numbers before claiming its operation was a social experiment exposing weak scrutiny in polling and media.
Newsweek retracted three articles and corrected others after the deception emerged, while polling databases removed the firm’s surveys.
The fake pollster's suggestion of a close Mississippi race happened to resemble the outcome, but it was based on invented data.
What Comes Next
The Mississippi result does not dismiss polling as useless, but it does demonstrate the danger of turning limited evidence, familiar voting history and uncertain turnout assumptions into confidence.
The next test for forecasters is whether postelection reviews explain how warning signs were weighed, when categories were reconsidered and how uncertainty will be communicated before the next unlikely candidate asks donors to believe a race can be won.
Newsweek*’s reporters and editors used Martyn, our AI assistant, to produce this story. Learn more about Martyn* here.
Contact Newsweek editors on this story: John Fitzpatrick and Geoffrey Rowland.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines