A few days ago, Donald Trump’s support among African American voters in the USC Dornsife/L.A. Times Daybreak tracking poll of the election appeared to shoot upward.
Since early summer, when the poll started, Trump’s support among black voters had been in the low single digits in the poll, as it is in most surveys. Suddenly, he seemed to be nearing 20%.
Some of Trump’s supporters cheered and began developing theories for why their candidate had finally started breaking through to black audiences. Outraged liberal critics of the poll denounced it anew.
And then, just as quickly as the line on the chart had turned upward, it turned back down. As of Wednesday, Trump’s black support in the poll is back to the single digits, near where it had been all along.
Detailed results from the USC Dornsife/LAT daily tracking poll >>
What happened is an object lesson in how not to read polls, particularly a daily tracking poll such as the Daybreak survey.
All polls are subject to random statistical noise. Tracking polls, because they take a sample every day, are particularly likely to jump around for no reason other than chance. That’s especially true with a small sub-group like African Americans, who make up about one-eighth of the electorate.
The change in Trump’s support was always well within the poll’s margin of error for black voters, meaning there was a good likelihood that what appeared to be a shift was just random. Now that the level of support has returned to where it was, that seems likely to have been what happened.
The lesson for poll watchers: Be wary of short-term fluctuations, particularly those involving subgroups. Take margins of error seriously. And don’t leap to conclusions until the evidence is solid.