Comment by tovej

Replication != validation, replication would be purely for experiments, which usually do not use that much causal language or attributions to theory. I'm talking about validating theory, and specifically neoclassical theory.

But still, that is interesting. Thank you for taking the time. Although you first source is a survey of expectations by authors on replicability over time, not an actual measurement of replicability.

Your second reference says there are historically more replication studies being published now than before (people trying to replicate other studies), not that more studies replicate.

As you might be aware of, the number of scientific publications in general is growing, so I would also expect the subset if replication studies to grow. This is not very surprising or meaningful, but interesting.

The third one is actually about reproducibility, the authors have taken open data sets published together with papers and see if they can produce the figures and final values in the papers from the original data sets. That is an entirely different thing.