Shaky foundations, parameters and assumptions arbitrary and designed to suit the story. I bet half of them have coding errors too
All structural models are complete BS
-
Shaky foundations, parameters and assumptions arbitrary and designed to suit the story. I bet half of them have coding errors too
It's hard to appreciate how true this is until you do structural work, and absolutely it's true: there are so many coding decisions, lots of potential for errors, almost never reviewed carefully. Structural is hopeless.
-
There is nothing worse than researchers who do not understand the limitations of what they do. With structural unfortunately we are still in the phase in which proponents truly believe they are doing something good because they do not understand the limitations. We went through data with cutesy identification strategies and we are now over it. Hopefully we will be over this naive approach to structural too quite soon
-
This. I took an IO course in grad school that required us to replicate some of the top papers. I was absolutely shocked at how shaky things were, results literally flip signs even when you make the slightest change to model setup. And yet everyone in the field pretends that the results are credible and robust. Complete joke.
Shaky foundations, parameters and assumptions arbitrary and designed to suit the story. I bet half of them have coding errors too
-
I think reduced form regressions even with million data points is sketchy. Even for short T, there could be numerous policy decisions affecting the intertemporal decision making other than the one you are focusing on. Government regulation and policy change pretty much all the time.