President Trump’s Approval Rating Is at a Near-Record Low. Here’s What to Know About the History of Those Numbers

More than a month into a partial government shutdown, Americans are apparently pretty unhappy with how President Donald Trump has been handling things. A new Associated Press-NORC Center for Public Affairs Research poll, released Wednesday, reports that Trump’s approval rating is 34%. That’s down from 42% a month ago, and close to the worst rating he’s received yet.

At such a fraught moment in government, many look to such polls for an indication of how decisions made in Washington are affecting the national mood from coast to coast. Perhaps unsurprisingly, their history started at another moment of great controversy about the role of the president and his office.

In the early 1930s, under FDR, the U.S. had undergone a massive increase in the size of the federal government. But though the Great Depression was still ongoing in late 1935, stocks were up and the nation seemed to have reached a level of calm. That fall, George H. Gallup, founder of the American Institute of Public Opinion, had released his very first set of political opinion polls, and they revealed that 60% of Americans felt New Deal programs were too expensive. And as confidence in the economy ticked up, the popularity of the president behind those radical programs was trending down.

By March 1936, TIME characterized Gallup polling data as “probably as accurate a sample of public sentiment as is available,” and presidential approval ratings were part of the story from the beginning.

Roosevelt knew it, too. With the upcoming 1936 election — in which he vied for a second term against Republican Governor of Kansas Alf Landon — seen by many as a referendum on the him and his agenda, he was just as interested as citizens were. It wasn’t that FDR was the first president to want that kind of report card. But until then, presidents had used a number of different (and not very scientific) ways to gauge what people thought of them. For example, they looked to newspaper editorials and letters to the editors, or they counted major labor strikes. But by the 1930s, American society had changed so much that the old ways didn’t really work, if they ever had.

There was a need to figure out how to take the temperature of the population. The people who figured out how to do so weren’t initially acting for political reasons. Rather, the need to gather public opinion evolved out of an ongoing Progressive Era trend toward the use of facts and figures in the service of social reform, says historian Peter A. Shulman, Professor of History at Case Western Reserve University and editor of Historical Opinion, a Twitter feed dedicated to historical polling data. After social reform came commercial usage, and then political polling.

“Cities had gotten bigger, Americans don’t know their neighbors anymore, people are speaking all different languages. In the late 19th century and early 20th century is the early development of modern social science in all these different related fields,” he explains. “The bigger picture was trying to get a sense of some kind of mass public, depending on what you’re trying to understand. Marketing surveys, trying to figure out what consumers wanted, were [part of] a relatively new field in the 1920s, with mass production, mass consumption, radio, newspapers. Inevitably that work would get applied to politics.”

Straw polls to predict elections came early on; Literary Digest, a general-interest weekly magazine founded in 1890, started what TIME called “the first nationwide sampling of public opinion,” and accurately called the 1920, 1924, 1928, 1932 presidential elections. But for the 1936 race, the publication incorrectly predicted a Landon win. One reason for the mistake was that the results skewed towards wealthier voters who were more likely to vote Republican, and “took no account of the low-income voters who had swung solidly behind the New Deal,” TIME has pointed out. “Anyone would guess that more Landon than Roosevelt voters were to be found at their 1932 addresses in 1936.” Cost-cutting measures affected the data-gathering process too, as TIME reported that Literary Digest mailed 10 million ballots, “half as many as in 1932.” (The magazine folded in 1938, and TIME inherited its subscriber database.)

Gallup, however, correctly called the winner of the 1936 election, having predicted in late October that Roosevelt would get 54% of the popular vote. From there, his influence on presidential approval ratings history continued, and expanded far beyond election season.

He is credited with pioneering “scientific” presidential approval ratings. In 1935, his polls asked, “Did you vote for Franklin Roosevelt in 1932?” and “Would you vote for him today?” He figured out a way to “sample a section of the public big enough to be statistically accurate, representative enough to include day-laborers, skilled workers, farmers, white-collar employees, millionaires, etc. in the same proportions in which they are found in the population at large,” as TIME put it that year. Roosevelt’s White House was interested, unsurprisingly, in what Gallup was doing. Roosevelt had his own pollster Emil Hurja, who is known for pioneering the job of a political polling consultant, and Hurja regularly consulted Gallup.

For the next few years, the wording of the poll varied as Gallup experimented with saying “in general” and “today” or “Are you for or against Roosevelt?” Since Harry Truman’s administration, Gallup’s standard wording for presidential job approval polling has been: “Do you approve or disapprove of the way [the President’s name] is handling his job as president?”

By 1948, a TIME cover story called Gallup the “Babe Ruth of the polling profession.”

By using the same wording over time, pollsters have been able to notice some trends. According to Lydia Saad, Director of U.S. Social Research at Gallup, “Usually presidents start off with high ratings, for obvious reasons: the goodwill brought about by their inauguration, benefit of the doubt and leeway at beginning. Then as things start happening, it will go down over next three years.” Trump, Saad says, was an exception: “He came in with little to no honeymoon.” (Gallup’s latest polling, which was conducted about two weeks into the shutdown, gives Trump a 37% approval rating.)

The data is also affected by factors such as what political scientists call the “rally ’round the flag” effect, as presidential approval ratings will spike during times of threat. (Most notably, George W. Bush’s 90% approval rating after September 11th, was the highest recorded in Gallup history. His father had enjoyed a similar bump during the collapse of the Soviet Union and the first Gulf War.) The State of the Union address, meanwhile, tends to have “little effect” on Americans’ views of the President, according to a recent Gallup analysis of the subject.

But a president’s personality can also have something to do with it. In 1990, TIME posited that the first President Bush’s high approval ratings were the result of him “keeping his head down” during his presidency. “In a slick piece of reverse psychology, he strives for underexposure: while most politicians crave attention, Bush made a conscious decision before his Inauguration to avoid appearing regularly on the nightly news,” the story noted. “He not only wants to lower expectations that a President can solve the nation’s problems but he also fears that his re-election will be more difficult if the public wearies of his visage in the first few years. ‘People get tired of seeing anybody on television,’ says a senior White House aide. So Bush stays on the margins of public consciousness, betting that in today’s peculiar politics, as in romance, absence makes the heart grow fonder.”

Trump, on the other hand, tends toward the opposite strategy, regularly tweeting updates about all sorts of things — not least his own approval ratings.