One of the most popular articles last week involved claims that polls showed Republicans had increased their support of President Trump. But a closer analysis of the data reveals that any increase in support was within the margin of error. So the polls couldn’t conclude that GOP support for President Trump had gone up or down.

When news organizations touted that “President Donald Trump’s approval rating with Republicans rose in a poll taken after his racist tweets on Sunday attacking four Democratic congresswomen of color,” I decided to investigate. These stories claim that Trump’s approval ratings rose “rose 5 percentage points to 72 percent from a similar poll conducted last week.”

So, I looked at the Reuters polls from July 10 and July 17. On July 10, 53 percent of Republican registered voters strongly approved of the President Trump, with 28 percent of these GOP voters “somewhat approving” of President Trump, with 2 percent of GOPers leaning towards approval. On July 17, 52 percent of Republican voters strongly approved the President, with 32 percent of the GOP somewhat approving, and 2 percent of Republicans leaning towards approval.

That means the President’s doing better with Republicans in the polls than what was reported in these articles. But that “net approval” was only three points (from 83 percent to 86 percent ), well within the 5.6 percent margin of error in the Reuters/Ipsos polls. Go look it up, and see for yourself.

Trump’s approval rating fell dramatically with Democrats and Independents. However, those were in the margins of error too. As the fivethirtyeight.org pollsters would say “It’s a bad use of polling.”

Additionally, Republicans from elected officials to key advisers worked overtime begging Trump to walk back the comments, or at least soften the tone. While most were leery of criticizing Trump directly, they did call the words a mistake or regrettable, as reported by numerous sites. That’s not what you do if those tweets are being well received by Republicans. President Trump’s approval ratings are also well below 50 percent in both Reuters/Ipsos polls.

At a minimum, we cannot conclude at the writing of this article that Republicans are more likely to support President Trump after those tweets, or are less likely to do so. But what we can conclude is that our polling is not what’s incorrect, but often our analysis of it.