Prediction Penance

by: Cullen Roche

John Cochrane wrote a follow-up post to his recent attack on Keynesian economics in which he clarified some of the views. I particularly liked this line:

The basic "trouble with macroeconomics," circa 2014, is about the same as the "trouble with physics," circa 1790. We know a lot. But there is so much we don't know.

Yes, we do know a lot. We can understand far more about the operational realities of our monetary system and the financial system than most economists would have us believe. The problem, however, is that many economists are so blinded by their politics that they refuse to look at the world through the prism of operational realities.

For instance:

The last 5 years have been particularly enlightening for macroeconomists because this period has shed light on people who are using flawed understandings of the world. All of the predictions I made during this time were based on my operational understandings of the way the monetary system. I didn't inject my personal politics into this. I simply studied things like QE and fiscal policy inside of the specific macro environment (a deleveraging) and tried to come up with high probability outcomes.

Now, maybe I was lucky, but one thing we know is that John Cochrane predicted a USD collapse, high inflation, rising interest rates and a potential debt crisis back in 2011. So while these predictions don't necessarily prove that my model of the world is right, they should shed significant doubt on the validity of his model.

I don't intend to sound like I am beating up on Cochrane's predictions from 2011. No, I am simply making a point about the importance of having a model that actually reflects our reality. I know we're all going to be wrong about things at times. I certainly make plenty of mistakes, but we should really embrace these bad predictions - especially when someone's model of the world predicts that we are at "the brink of disaster". How else can we learn from mistakes and make progress if we continue to rely on models that produce hugely erroneous predictions?