I think local predictions are hard for any school to make accurately in economics, but broad swath stuff is doable. So, to have a decent shot at a good theory,

I think local predictions are hard for any school to make accurately in economics, but broad swath stuff is doable. So, to have a decent shot at a good theory, you need to figure out what the most important drivers are in your situation of interest. In business cycle theory, I think it is the action of the interest rate over time which is the most important thing that gets short shrift in the main, and which drives the dynamics I speak about above. I can make a regression say anything, but I predict the fed continues to operate, and the status quo of wealth flow from poor to rich continues. Now it can be modified by laws, and certainly will change with time, but the dynamics are there. You can superimpose other things on top and the results will vary. You can then make a model that focuses on one thing or another, and evaluate it statistically, and again, results will vary. This is some of how we ended up with a pluralist economics today. I wrote up some other stuff about my anecdotal experience with modeling and searching for "good governance" in grad school, and how I disagreed with methodology of the papers I was reading, before abandoning the enterprise for a return to physical engineering, but I guess this is my chief point. Methodology designs in what you want to see. Economics is not a controlled science (I mean mostly the studies can have no actual control group), and therein festers a great rub. Good fun learning the statistical analysis methods though.