Developer strength matters more than most other factors. A strong developer will run circles around a weak one even if the former is programming in assembler and the latter in Python. Thus studying variables such as programming language is difficult and costly because the variance is enormous. Therefore it is not done and a result like web developers are on average more productive in Python than in assembly is probably not that interesting anyway. Thus we are left with "truths" there is no scientific basis for.