The better critique is that math requires homogenous data points.
This critique is similarly wrong. There's a field called "machine learning" - you might have heard of it - which often uses inhomogeneous data points.
So doing statistical studies on a question like "do charter schools work" that treats all these implementations as identical, is often not very useful for actually learning something about policy.
So instead build a model that understands the important factors which make charter schools work/fail. Then use it.
If you can't do this, don't pretend you have anything to say about whether I should implement charter schools.
This critique is similarly wrong. There's a field called "machine learning" - you might have heard of it - which often uses inhomogeneous data points.
So doing statistical studies on a question like "do charter schools work" that treats all these implementations as identical, is often not very useful for actually learning something about policy.
So instead build a model that understands the important factors which make charter schools work/fail. Then use it.
If you can't do this, don't pretend you have anything to say about whether I should implement charter schools.