Opinion polling, for all of its problems, remains a reasonably good way to answer very simple yes-or-no questions, such as determining who will win an election or whether or not people approve of the president. So long as we take it as a probabilistic exercise rather than a precise science, we can learn a lot from polling. We can also gain some ideas about why people vote the way they do by looking at, say, a president’s general approval on areas such as the economy, immigration, or foreign affairs.

The further we get from those elementary sorts of polls, however, the more malleable polling becomes. Consider two polls released in the past few days. YouGov asked 4,309 U.S. adults on June 10, “Do you approve or disapprove of deploying Marines to the Los Angeles area to respond to protests over the federal government’s immigration enforcement?” Only 34 percent approved, 47 percent disapproved, and 19 percent said that they were not sure. On June 9–10, the Republican-leaning pollster Insider Advantage asked 1,000 likely voters, “What is your opinion of President Trump’s decision in recent days to deploy National Guard and federal military in downtown Los Angeles?” That yielded 59 percent approving, 39 percent disapproving, and 2 percent undecided.

So, either about a third of the public approves of the deployment — or nearly 60 percent do. What gives? It’s always helpful to have more polls to measure the answer, but these polls were taken on different dates, with different sample sizes, of different pools of people, asking different questions. YouGov talked to adults in general, while Insider Advantage polled likely voters. That means YouGov was polling a lot more people who follow the news less closely (hence, the much higher proportion of “not sure” answers), may not be registered voters, and may not even be citizens; on the other hand, it means that Insider Advantage is doing more massaging of the sample to narrow it to likely voters, which is typically a hard thing to pin down a year and a half before the next federal election. The two pollsters also loaded the question in different ways by the way they phrased it. YouGov asked specifically about deploying the Marines, not “the National Guard and federal military”; more pointedly, it framed the question as a deployment “to respond to protests,” which prods the listener to think of this as quashing protests rather than restoring order against rioters. Insider Advantage did no such priming of the respondents, leaving them to fall back on whatever they already knew or believed about why the military was being called out.