The world can be a very confusing place. Cause and effect relationships abound, but are often hard to tease out of data. Never mind the many pitfalls to understanding that distort our perceptions of data and bias our understandings, some systems are so complex, with so many variables that impact each other and still other variables in varying degrees, until predicting aggregate system behavior through modeling is nigh well impossible. The economic system of a nation as vast and complicated as the US is one such system, as Caroline Baum points out at Bloomberg.com today, describing the econometric model the Fed uses in making its forecasts or in explaining the past:
That would be the same model that failed to grasp that mortgage loans made during a period of exceptionally low interest rates by lenders with no skin in the game might not be repaid, putting major financial institutions on the brink of insolvency; the same model that failed to understand how new and exotic derivatives based on these mortgages would perform; the same model that failed to see the millions of jobs that would be lost if housing and credit bubbles were allowed to inflate until they burst; and the same model that predicted an unemployment rate of 8.8 percent in the fourth quarter of 2010 without the enactment of a fiscal stimulus. (It was 9.6 percent with it.)
Why do policy makers persist in perpetrating this fantasy, in asserting something that can’t be proven? Any econometrician will tell you “the error bands are huge and consistent with almost any result one might imagine,” said Bob Eisenbeis, chief monetary economist at Cumberland Advisors, in a Jan. 11 commentary on Yellen’s speech.
Thus the Fed’s econometric model can’t even predict the past. The range of error roughly corresponds to any imaginable result. And we listen to these people who rely on these models why?
Believing that an economic system so vastly complicated, with literally trillions and trillions of discrete effect-causing decisions happening every second of every day, can be modeled with sufficient accuracy to provide actionable predictions is tantamount to believing that human thoughts and behaviors can be predicted by examining the relationships and activity of the trillions of individual neurons in the brain. Try as we might, in both instances we’ve never even come close. If we assume that all causes have effects–the basis for rational inquiry in any endeavor–then we theoretically ought to be able to create models of any dynamic system and achieve some measure of confidence that our predictions are valid. Yet, as a practical matter, complexity wins. Our meager ability to create computing devices (all the supercomputers in all the world don’t have the computational capacity of even one living cell, brain or otherwise) limits our ability to model complex, dynamic systems such that our models account for all the variables and inputs that might affect them. Neurons in the brain are so intricately connected and interwoven, conscious thought is a mysterious, emergent quality that we may never completely understand. It is hubris of the sort that yielded investment bank models that predicted housing prices would never go down (before the crash) to believe otherwise.
There was a time when people believed that the alignment of the stars in the heavens on the day of one’s birth determined the outcome of one’s life. We scoff at such irrational and superstitious behavior today. Everyone knows that the twinkling stars in the heavens are actually enormous suns, many millions of light years away, and that their gravitational pull, among other possible effects, is so infinitely small as to be virtually unmeasurable. Yet we think it rational to believe in our ability to model complex systems, even as we’ve seen time and again that our models inevitably fail. Belief, it seems, did not die when astronomy replaced astrology. It was just transferred to the realm of pseudo-science. The essence of rationality is understanding what can be known and what can’t be known. So far as systemic economic behavior goes, what we know is far less than what we believe we know. Which ties in nicely w/ the idea presented in a previous column by Baum that politicians can’t grow an economy. If we don’t even know for sure how it works, how can we know how to make it grow?