# Trump's approval rating is roughly 36%. Could the polls just be wrong?

The polls are almost certainly somewhat wrong, but not by much. (Some are politically biased as well, but that’s a separate issue and the biases are often in favor of the GOP.) All polls use statistics to quantify how much wrong they could be. When a poll reports “Trump’s support is at 37% plus or minus 2.0 percentage points” (as did the recent Reuters/Ipsos poll), then that true value is between 39%–35% with the most likely value being 37%. Some polls or news reports don’t report the error range, which is unfortunate and deprives you of useful information.

If you want to know how that error value is calculated and what affects it, you’ll want to study statistics and understand how that works in some detail. More detail than I can answer in this question.

## 12 Replies to “Trump's approval rating is roughly 36%. Could the polls just be wrong?”

1. Angela Stockton says:

The polls were not wrong. Every poll has a margin of error usually of between 2.5% to 3.5%, and they still do. Consequently the poll you are quoting is really showing an approval rating based on the margin of error from the lower to upper thirties. The last polls before election day showed HRC with a 1% lead but if you take the margin of error into consideration it was a statistical tie. Trump won by 78000 total votes in the states of Wisconsin, Pennsylvania, and Michigan and Clinton won the popular vote by about 2.9 Million votes. That means that the poll was dead right based on the margin of error but Trump wins an Electoral College victory based on about .0006% of the vote. Polls if conducted scientifically are a flawed snapshot in time and if not would not claim to have that margin of error. There may have been some Bradley Effect in which people voted for Trump while denying it but still the polls are correct within the parameters that they are designed.

2. Angela Stockton says:

The problem with pre-election polls is that some of them were taken after James Comey announced his discovery of “new” Hillary Clinton emails, but before he announced that they weren’t new after all. Also, a certain percentage of American voters are notorious for not making up their minds until they’re actually in the voting booth.

Anyway, the polls were not wrong that more Americans preferred Hillary to Trump. They were wrong only about how the distribution of Hillary voters would affect the electoral vote.

Every election that proves a poll wrong is a lesson in how not to conduct the next poll. For instance, The Literary Digest learned in 1936 that a mail-in poll is not scientific, and that its generally conservative readers, who tended to not like Franklin Roosevelt, were not typical of all Americans.

For that reason, the polls that measure Trump’s approval are probably trustworthy.

3. Angela Stockton says:

The polls are probably incorrect. Take note that Trump's current approval rating is still about 10 points higher than Clinton's approval rating. People still despise her a lot more than they do Trump. With good reason.

As far as the total votes indicating the accuracy of the polls … that's completely off base. The only place where the polls matched up to the results was in California, which Clinton won by more than 4 million votes. Her total votes in the election were slightly less than 3 million votes more than Trump.

That means that Trump won the popular vote in the other 49 states by more than a million votes. That pretty much makes the pollsters and pundits more wrong than previously considered. Are you gonna trust them now?

4. Clare Burke says:

Here’s the problem with the approval rating poll.

325 million population

210 million register voters.

65 million voted for Clinton (19% of total population)

62 million voted for Trump (18% of total population)

This poll only asks adults, 18 and older if they approve of the President. The population of 18 and older is roughly 260 million, which make the % vote for each candidate, among 18 and older, 25% and 23.8%.

According to the polls, more people approve of Trump than voted for him.

This is the problem w/polling and it’s why the polls completely goofed this last election.

They would do better to actually explain what they are doing instead of throwing a number out there, which doesn’t help anyone.

The polls claimed Clinton would win in a landslide and the polls now say Trump’s approval rating is in the toilet.

Are they accurate?

That’s the question.

5. Matt Hastings says:

Polls are not meant to be right or wrong; they are inherently probabilistic meaning they give you chances of certain outcome. And they were that far off, most national polls taken after the Comey letter news release had the actual percentages of vote well within their respective margin of error.

And job approval ratings are less complicated and more accurate than election polls since there’s no secondary decision to be made after the survey; any changes after the response would just be capture in the next period of polling.

6. Clare Burke says:

If only half the voters voted, and all of 36% vote, that would reflect an actual increase in approval since the election.

7. Charlie Lee says:

Polls are rarely wrong. The in of the data by humans can be misleading though. Hillary did win by the 3 million but when the numbers are divided up by the electoral college she lost. Random statistical sampling is surprisingly accurate.

8. Lawrence Hughes says:

The thing about polls is that they are always limited by their sample size. Who they asks to poll, who actually takes the time to fill out the poll, and how big of a sample size that they have. Thus even the most thoroughly well done poll is an estimate at best. To get a true, 100% accurate approval rating, we would have to ask every single eligible voter to fill out the poll. And no one has the time or resources to force everyone in America to fill out a poll. So while it is possible that a majority of Trump supporters have decided not to fill out the poll, or that Anti-Trump Supports aren’t answering polls, generally when enough independent pollsters share their data together, you get a realistic estimate of what the actual approval rating is over any one source’s best guess. They aren’t going to be off by a factor of 10.

TL;DR: Yes they could be wrong, but this is why we fact-check.

9. George Peterson says:

The polls were not really wrong. In the days leading up to the election, 538 gave him between a 20 and 30 percent chance of winning. Not good odds, but not terrible.

10. Michael L. Sutton says:

Polls mean nothing. They're only polling in certain places, at certain times. It's not like they polled millions of people. It works both way…

11. Clare Burke says:

Polls, any polls, will always say what you want them to say.

12. Ed Kelly says:

Of course they could, remember the election in November?

I choose to believe they are wrong in giving him 36%; I believe it is more like 20%.