One Poll Is a Rumor. The Whole Field Is the Fact.

The dangerous clue is the one that confirms itself

I built crossword puzzles for a living for thirty years. Here is the first thing that job teaches you, and it took me longer to learn than I would like to admit. The most dangerous entry in a grid is not the one you cannot solve. It is the one you are sure of. You write it in fast, in ink, and you move on, and it is wrong, and it quietly ruins everything that crosses it.

The only defense is the crossings. You do not get to trust an answer because it feels right. You earn the right to trust it when the letters coming down through it agree with the letters going across. A single entry is a guess wearing a confident face. The grid is the proof.

I thought about that all week, because the country just watched the same lesson play out in public, in numbers, about the president.

Eight answers to the same question

Between July 16 and July 30, eight national pollsters asked American adults whether they approve of the job President Trump is doing. Read them side by side. Quinnipiac said 32 percent (Quinnipiac University, July 29, 2026). The Associated Press and NORC said 33 (AP-NORC, July 30, 2026). CNN, polling with SSRS, said 34 (CNN, SSRS, July 29, 2026). Pew Research said 34 (Pew Research Center, July 23, 2026). The Washington Post and Ipsos said 37 (Washington Post, Ipsos, July 16, 2026). CBS News and YouGov said 39 (CBS News, YouGov, July 26, 2026). Emerson College said 39 (Emerson College, July 23, 2026). CNBC said 40 (CNBC, July 17, 2026).

Eight points of spread, from 32 to 40. If you wanted to write a story that the president had collapsed, you took the 32 and ran. If you wanted to write that he was holding, you took the 40. Both numbers are real. Both were published by serious people. And a reader who grabs either one alone is doing exactly what I did as a young constructor when I inked in the confident wrong answer. He is trusting a clue he has not crossed.

The crossings

So cross them. The way you check a field of polls is you take the average, because the average is where the individual house leans cancel out and the shared signal survives. On July 30 the Silver Bulletin tracker put the president at 38.2 percent approval against 58.8 percent disapproval (Silver Bulletin, July 30, 2026). That is the crossing letter. It does not match the alarmists at 32 or the calmest read at 40. It sits in between, and it is the number I would trust over any one poll in that list, because it is the only one that has been checked against the others.

Thirty-eight is not a good number for an incumbent, and I am not going to pretend it is. But it is a fact you can stand on, and 32 and 40 are not. They are the edges of the honest uncertainty. The middle is the answer.

A move is not the same as a mood

Here is the second discipline, and it is the harder one. The average did not just sit low this month. It moved. On July 12 the same tracker had the president at a net of minus 16.4 points. By July 30 it was minus 20.6 (Silver Bulletin, July 30, 2026). Four points in eighteen days is a real shift, not the random wobble of a single survey, and when a whole field moves together you look for the thing that pushed it.

This time the thing is not a mystery. Approval of the American military action against Iran fell from 44 percent in March to 39 percent by late July in CBS and YouGov polling, and Quinnipiac found 60 percent of voters disapproving of that conflict against 34 percent approving (CBS News, YouGov, July 26, 2026; Quinnipiac University, July 29, 2026). A war the country is unsure it wanted is heavy to carry in a number that measures the mood of a month. It got carried. That is an honest read of why the field moved, and it is a different statement than saying the presidency has failed. A mood is not a verdict.

The receipts on how far numbers fall

Now hold the whole thing in proportion, which is the part the loudest voices skip. A president in the low 30s is unpopular. He is nowhere near the floor of what this office has survived. Harry Truman fell to 22 percent in February 1952 (Gallup). Richard Nixon sat at 24 percent the month he resigned (Gallup). George W. Bush touched 25 percent in the fall of 2008 (Gallup). The 34 percent CNN measured this week is not a historic bottom. It is the same mark President Trump hit once before, in January 2021, and he was elected to this office again four years later (CNN, SSRS, July 29, 2026; Gallup).

I am not saying that to comfort anyone or to worry anyone. I am saying it because it is on the record, and the record is the only thing I sell. Numbers like these have been reached before, by presidents of both parties, and the country kept its footing every time. That is not a partisan sentence. It is just what the eighty years of Gallup data say when you read the whole column instead of the last cell.

Why I do not rank a president by his polls

People ask why US Political Rank never scores a president by approval. This week is the answer. Approval measures how a month felt. It is a rumor until you cross it, a snapshot even once you do, and it moves on a war or a jobs report before the ink is dry. It is a real and useful thing, and it is not the record.

The record is what a term actually produced, and that is what we grade. When we rank presidents by economic record, we use full-term averages of growth and inflation from the federal statistical agencies, numbers that do not care whether the man was cheered or booed while he governed. And when a pollster's figure surprises you, remember the pollsters keep score against each other too, with track records that reward the ones who get it right (see pollster accuracy rankings).

Read the field, not the clue. Take the average, weigh what moved it, and remember how far this office has fallen before and climbed back. Do that and you will never be the constructor who inked in the confident wrong answer. You will be the one who checked the crossings first.