By Jim Ellis — Monday, March 3, 2025
Polling
President Donald Trump / Photo by Gage Skidmore
The first set of contradictory surveys pertains to recent job approval ratings for President Donald Trump. The two polls in question produced diametrically opposed results when asking the same question during the same time period.
The American Research Group tested the Trump favorability rating over the Feb. 17-20 period (1,004 registered US voters; live interview) and found the President with an upside-down ratio, 43:51 percent favorable to unfavorable, a negative spread of eight percentage points.
Within the same time frame, Harvard University (HarrisX/Harris Poll; Feb. 19-20; 2,443 registered US voters; online) also tested the American registered voter electorate but found a completely different favorability ratio. According to the Harvard data, Trump has a positive rating spread of nine percentage points, 52:43 percent favorable to unfavorable.
Therefore, the two professionally conducted surveys, executed within the same period (Feb. 17-20), and targeting a like audience (registered voters) for the same purpose (testing presidential job approval) arrive at completely different responses.
It is probable that the more positive Harvard poll is closer to the accuracy mark, as another surveyor, the Morning Consult data organization, conducted a national tracking poll within the same late February period. The MC study (Feb. 21-24; 2,225 registered US voters; online) projects Trump with a plus-3 favorable job approval ratio, 50:47 percent.
The large disparity found within these and other similar conducted surveys over the weeks since President Trump began his second term on Jan. 20, suggest certain polling flaws or that the public is displaying inconsistent hot and cold tendencies toward the former and current chief executive. Even the latter potential conclusion, however, is an improvement when compared to the President’s first term standing when his approval ratings were uniformly negative.
A newly released campaign poll should be ignored, but not because the data results are necessarily wrong. The reason is one of the tested principals has definitively stated he is not running for the Senate.
The campaign in question is the 2026 Massachusetts Senate race featuring incumbent Sen. Ed Markey (D), who has been in Congress since 1976 counting his long career in the House, and former two-term Gov. Charlie Baker (R). Baker, now president of the National Collegiate Athletic Association (NCAA), has repeatedly said that he will not run for the Senate.
The survey, from the University of Massachusetts (YouGov; Feb. 14-20; 700 Massachusetts adults; online), posts Sen. Markey to only a two-point edge over ex-Gov. Baker, 35-33 percent. Aside from testing a candidate who is not running, the sampling universe consists of adults and not likely or even registered voters. Therefore, the poll results, which also appear to under-count each man’s support, are not useful.
Another questionable statewide survey, but one that does monitor candidates who are running, tests the Virginia electorate regarding their open 2025 Governor’s campaign.
The Roanoke College Institute for Policy and Opinion Research, which has been known in the past for releasing certain unusual survey results that were later proven as anomalies, fielded their study of 690 likely Virginia general election voters over the Feb. 17-20 period. The Roanoke ballot test sees former Rep. Abigail Spanberger (D) leading Lt. Gov. Winsome Earle Sears (R) by a 39-24 percent spread, which appears to underestimate not only Sears’ support, but also Spanberger’s.
Furthermore, their ballot test results are inconsistent with the other five Virginia Governor polls conducted and publicized since the 2024 election. The latter surveys, from five different pollsters, cast Spanberger and Sears in a dead heat (co/efficient survey research firm), Spanberger plus-1 (Emerson College), plus-3 (Mason-Dixon Polling & Strategy), plus-5 (Christopher Newport University), and Spanberger plus-10 (Virginia Commonwealth University).
While these polls still provide some useful information, those currently testing political campaigns produce results so early in the election cycle that typically fail to properly capture the studied candidates’ accurate standing in reference to a future final result. This is because polls can only test one point in time.
Therefore, once we see campaigns beginning to spend voter contact money to deliver their message and project a theme, the ballot test results can be more seriously considered and analyzed.