The Hoover Institution at Stanford University has always seemed to me a conservative think tank, dominated by a 285-foot tower honoring a one-term president who was born in 1874. Students are drawn to Stanford not by Hoover but by the university’s involvement in the thrilling innovations of the 21st century.

The Hoover Institution scholars who research education, however, impress me. I can’t call them conservative. That adjective doesn’t apply to arguments over schools. They are unconventional thinkers, and sometimes troublemakers, as is apparent from an education policy analysis they just released.

Their paper, “The Unavoidable: Tomorrow’s Teacher Compensation,” paints the clearest picture I have seen of U.S. schools’ troubles and offers radical solutions. They want, for instance, to take some of the money going to teacher retirement funds and use it to raise salaries, particularly for the best teachers.

I will describe the research first, and next week the suggested cure. What they propose sounds unlikely, certain to inspire fierce opposition. But parts of the country are experimenting with bits and pieces of what they want.

American schools could be better. According to the most accepted measuring stick, the National Assessment of Educational Progress (NAEP, pronounced “nape”), fourth-graders have made gains, but they disappear later. The 2019 NAEP report said eighth-grade math scores were stagnant and reading scores falling. Twelfth-grade scores have been essentially flat since the 1970s.

The Hoover researchers, often at odds with teachers unions, agree with them on the need for better pay. Teachers on average earn 22 percent less than people with comparable educational attainments, their analysis said. Increasing compensation provides the best leverage available to improve student achievement, the analysis said.

The policy analysis was written by economist Eric A. Hanushek, a prolific senior fellow at Hoover who is an expert on school spending. It is based on background papers by Maria D. Fitzpatrick, Marguerite Roza, Steven G. Rivkin, Ben Ost, Andrew Morgan and Minh Nguyen and funded by the new Hoover Education Success Initiative.

From 2000 to 2016, entry salaries for teachers adjusted for inflation were unchanged and average salaries for all teachers declined slightly. Expenditures per student in 2016 were $13,139, up 35 percent from 1990 when adjusted for inflation.

“For the century before 2008, per-pupil expenditures rose steadily, regardless of any other pressures on public budgets,” the analysis said. But growth sputtered. “Instead, it appears increasingly necessary to trade ineffective old programs for any new programs,” it said.

School board members across the land will nod at the analysis’s conclusion that a key limit on spending in the classroom “is that financial pressures of pension costs and retiree health costs are rising and are being passed along to districts.” School boards prefer to raise pensions rather than pay because the financial hit doesn’t come until later. The current unfunded liability of teacher pensions for the United States is estimated at more than $1 trillion.

States try to improve student achievement, it said, but “we have found that many programs, initiatives, interventions and reforms simply do not deliver the promised gains.” Requiring teachers to get more training and reducing class size have not worked well. Teachers tend not to improve after their first few years on the job. Drawing the best people to schools and keeping them there is effective, but that means they have to be paid more.

Hanushek has been seen as a leading advocate of rating teachers on how well their students do on standardized tests. But he told me that “one would never want to use just test scores for evaluations.” His analysis suggests looking not only at scores but also at classroom observations, teacher attendance and student surveys.

The key, he told me, is to have systems that differentiate seriously among teachers on effectiveness, rather than the usual practice of principals keeping the peace by giving high marks to 95 percent of their people. Then, schools must find ways to reward the best teachers with better salaries, the analysis said. It suggested diverting some money that would otherwise increase pensions to paying teachers more while they are still in the classroom.

I will discuss next week how the Hoover people propose to do that. I predict it won’t work, but I am a conventional thinker. Their reasons for saying I am wrong are interesting.