No one knows if Donald Trump’s 34 felony convictions will move many Americans when they vote in five months. Nevertheless, the early surveys probably cheered Joe Biden’s White House: A Reuters/Ipsos post-verdict poll showed the president now leading Trump by 2 points (41 to 39), compared to the pre-verdict RealClearPolitics average with Trump ahead by 2 points (42 to 40). (Robert Kennedy Jr. had 10 percent.)
These results tell us nothing about what will happen in November. Both findings are within their margins of error, so statistically, the results equally could have shown Biden up 2 points before the verdict and Trump up 2 points post-verdict. Let’s also set aside the technical issues that cast doubt that these surveys reflect which candidate is favored today by those who will vote in November (These problems are well documented in the Washington Monthly here and here).
Beyond all that, history shows that early June polls bear no reasonable relationship to general election outcomes, even the Gallup polls that are as close to an industry standard as we have.
Incumbent presidents have run for reelection seven times since 1980. While 2 percentage points currently separate Trump and Biden, the early June Gallup polls in those seven reelect cases differed from the ultimate election results by an average of 9 percentage points, including three cases in which the early June leader lost in November.
In 1980, when Ronald Reagan challenged President Jimmy Carter, the Gallup poll in early June showed Carter leading Reagan by 7 points (and ahead of third-party candidate John Anderson by 18 points). In the end, Reagan beat Carter by 10 points, so the June polls were off by 17. Four years later, the early June poll found President Reagan leading Walter Mondale by 9 points. Yet in November, Reagan’s margin was 18 points, so the June poll was off by 9 points.
The next race involving an incumbent came in 1992 when President George H. W. Bush was six points ahead of Bill Clinton in June—and Ross Perot was leading Bush by eight points and Clinton by 14 points. Clinton beat Bush by 6 points and Perot by 19 percent by election day, so the early June poll was wrong about the winner and the runner-up. And based on actual ballots, the June survey was off by 12 points on Bush v. Clinton (and 38 points on Clinton v. Perot).
Four years later, in early June, President Clinton led Bob Dole by 16 points (and Perot by 32 points). By November, Clinton beat Dole by 8 points (and thrashed Perot by 41 points), so the June survey was off by eight points on Clinton v. Dole.
The next incumbent up for reelection was President George W. Bush in 2004. In the early June Gallup poll, he trailed John Kerry by 6 points. Since Bush ultimately won by 3 points, the June poll was off by 9 points.
The outlier was the Obama-Romney race in 2012: President Obama led Mitt Romney by 3 points in June and won reelection by 4 points in November. Eight years later, in mid-June 2020, President Trump trailed Biden by 10 points—and in the final balloting, Biden beat Trump by 4 points, so the June Gallup poll was 6 percentage points off.
How did a top pollster, Gallup, compile such a record? To be fair, the company never claims that its June polls predict November election results. The problem is that politicians and pundits read far too much into the early survey snapshots. On top of the technical problems that can distort all political surveys, many people change their minds about many things over any five-month period, let alone when there’s a general election barrage of earned and paid media.
Yet, if an investor’s record over five months was as off the mark as the last 44 years of early June polls were at predicting November results, that investor could end up bankrupt. The lesson for this year’s election is don’t bet anything based on this week’s post-verdict polls.
