I believe we should do a better job of grading economists on how accurate their predictions are. I think we should create a central registry for economists to make predictions about individual statistics in the future. Once the prediction is made, it becomes permanent for everyone to see.
Taking this idea a step further, there could be a standard test for macroeconomists to predict a collection of about 10 statistics for an economy. Economists could fill out the statistics like a March Madness bracket. The weighted sum of errors for all the predictions would be complied to determine a numerical score for the macroeconomist. This could be some sort of competitive game with an annual winner based on predictions made five or ten years earlier.
What tools should a macroeconomist use for predictions? I think the best predictions in the short term would be based on mathematical models but predictions further into the future would rely more on judgment calls. Predictions far into the future (such as ten or thirty years) would require knowledge about new technologies, future political institutions, future consumer demand and the evolution of society. This is why macroeconomics is so difficult.
I think this test would reveal which theoretical systems are most effective for macroeconomics. Maybe economists who specialize in economic history would perform the best. Maybe the econometricians would perform the best. I'm not sure who would produce the most accurate predictions, but I think this would be a productive exercise for macroeconomists.