Harvard researchers found that even small to mid-sized media outlets can have a dramatic effect on the national conversation about public policy issues, in a unique study in which outlets agreed to publish certain stories on certain days so their effects on social media could be measured.

“For several hundred years, scholars have tried to measure the influence of the media. Most people think it is influential, but measuring this influence rigorously with randomized experiments has until now been impossible,’’ professor Gary King said in a university statement. “Our findings suggest that the effect of the media is surprisingly large.’’

The study was published in the journal Science.

The study found that if just three outlets, with an average circulation of about 50,000, wrote at the same time about an issue — jobs, the environment, or immigration, for example — discussion of the topic on Twitter rose by more than 62 percent in the following days, and opinions could be swayed several percentage points.

King said in a telephone interview that it looked like the effect would be even greater with bigger news outlets. But he said that journalists at small outlets still had a “pretty amazing’’ effect.

“Any journalist with any kind of audience has an incredible power,’’ he said. “[When people say] journalism is not just another job, gosh, it’s really true.’’

King said the researchers used technology from Harvard-based startup Crimson Hexagon, which he cofounded, to evaluate the meaning in posts as researchers combed through 500 million tweets a day. Researchers examined the data for the weeks that test stories were released. They also looked at comparison weeks when no stories were released.

He said the effect of the test stories on social media was the same, regardless of factors such as users’ gender, party affiliation, or geographical area.

“Even though we may all be in our own filter bubbles, we all join the national conversation on a similar topic at a similar time,’’ he said.

Conducted by King and former students Benjamin Schneer, now an assistant professor at Florida State University, and Ariel White, now an assistant professor at MIT, the study took more than five years of work.

Similar research had failed in the past because of journalists not wanting to be told what to report and when, the university said.

Forty-eight outlets helped researchers in their efforts, King said. Thirty-three of them actually engaged in the testing, teaming up in groups of two to five outlets for 35 different tests, King said.

King said it took years to navigate the “impossible task’’ of reconciling journalists’ desire for control over the content and timing of stories and the scientists’ need for the same control to conduct their study.

“We worked really hard talking to journalists,’’ he said, to find a way to cross the “unbridgeable chasm.’’

In the end, researchers were able to pick groups of media outlets, policy areas, story angles, and the timing of publication.

But, at the same time, the outlets working with the researchers could opt out at a number of stages — not joining the group or dropping out if they wanted to write a story with a different angle or publish it at a different time.

“The journalists can say they had total control, and we can say we had total control,’’ King said.

Martin Finucane can be reached at martin.finucane@globe.com