There was a good essay (somewhere… I forgot where) that argued that progress in science originates less from Kuhnian revolutions or shifts in methodology than from successes in defining homogenous objects. That is, genetically identical mice living in labs have a lot in common with each other, so we’ve made a lot of progress in genetics thanks to genetically-identical mice. But genetically-identical mice living in labs aren’t really all that much like people, so their utility in understanding human disease is constrained. Similarly, enzyme kinetics of a given protein and substrate is pretty much the same one time and the next, so we make progress defining and understanding it. But we don’t have a whole bunch of identical Earth’s atmospheres to mess with and manipulate, so climate is a fundamentally harder problem.
One large and related advantage of natural science over social science in advancing our understanding of the world is not just the relative durability of natural science theories but the relative constancy of those theories’ parameters, the constancy of the constants. The speed of light and the Boltzmann constant and, yes, the estimated heat absorption of a CO2 molecule have all changed less since 1900 than the elasticity of labor demand in low-wage markets changes from one NBER paper to another.
Apart from the Gap and related test score epiphenomena, it’s hard to know what enduring quantitative phenomena in education we are even trying to describe, what constants we are even trying to calculate, and if researchers even believe they exist.