The Fed's Modeling Mistakes
Last week, we shared empirical evidence that the Federal Reserve, employing the crème de la crème of technocratic expertise, has consistently failed to produce accurate forecasts of the macroeconomic future. This week we will discuss how much of this inadequate forecasting stems from fundamentally flawed economic theory.
In 2010, Allen Meltzer, a professor of political economy at Carnegie Mellon who sadly passed away this past May, completed his colossal work A History of the Federal Reserve after years of strenuous research. This epic history in two volumes and over two thousand pages repeatedly takes aim at Fed policies, theories, and forecasting methodologies that, according to Meltzer, have proven erroneous again and again despite their continued embrace by the technocracy. He asserts that one of the roots of these repeatedly poor performances is a theory central to modern macro orthodoxy, taught in essentially every undergraduate economics course throughout the nation: the Phillips curve.
The Phillips curve is a theoretical framework developed by A.W.H. Phillips in the late 1950s that argues inflation and unemployment tend to be negatively correlated and that a monetary authority could trade lower inflation for lower unemployment and vice versa. In an article published in the Cato Journal in 2010 summarizing the general themes of his magnum opus, Meltzer asserted that the continued use of the Phillips curve by the Fed to forecast inflation has led to grossly inaccurate predictions of future inflation. This squares well with Chang and Hanson’s findings from last week’s piece, which showed that while all official FOMC forecasts lack predictive accuracy, those focused on inflation tend to be particularly erroneous.
Meltzer explained that the problem with relying on the Phillips curve in crafting inflation forecasts and monetary policy is quite simple: the theory is wrong. The supposedly negative correlation between inflation and unemployment is not a permanent phenomenon of the macroeconomic universe but instead a relationship that holds true only in rare and unique periods in US economic history. Milton Friedman’s rational expectations model nuanced the theory by emphasizing that the Phillips curve may hold in the short run, but in the long run labor market participants would factor in changes in inflationary expectations such that they would no longer generate changes in unemployment.
Empirical evidence supports Meltzer’s and Friedman’s assertions. In their Q2 2017 review and outlook report, Hoisington Investment Management analyzed the relationship between changes in real wages and unemployment between 1965 and 2016. The resulting fitted trend line was not statistically significant, suggesting that the Phillips curve is not valid.
Figure 1: Real Average Hourly Earnings (YOY % change) vs. Unemployment Rate, 1965–2016 Monthly
Source: Hoisington Investment Management, 2017
While Friedman’s theory gathered great traction in academia, the Fed never abandoned the Phillips curve completely in forecasting inflation and determining policy. Instead, they attempted to improve the curve based on Friedman’s findings, ultimately leading to what is known today as the new Keynesian Phillips curve (NKPC). The NKPC, building off of Friedman’s ideas, asserts that, while under a flexible price regime (in the long-run) the Phillips curve relationship does not hold, in a universe of price rigidity (the short-run) it in fact does for several reasons. While the NKPC is widely accepted as a theoretical improvement upon the original Phillips curve, it continues to have the same issue: the real-world data does not fit the theory.
Even the Fed itself acknowledges this. The San Francisco Fed, in a 2007 report authored by senior economist Richard Dennis, admits that, “An array of papers has shown that the NKPC is unable to match the time series properties of aggregate inflation, failing to capture inflation’s persistence, overstating the role of expectations in price-setting, and implying what many believe to be excessive price rigidity. These inconsistencies between the model and the data are important, not least because much of our intuition for what constitutes good monetary policy has been built up using models in which the NKPC is central. A model that cannot satisfactorily explain why inflation is persistent is of doubtful value for forecasting.” While it asserts that researchers are exploring new ways of improving the NKPC, the report concludes that no viable alternative has been found to reliably explain the relationship between inflation and unemployment.
Thus, the Fed is left with a theoretical model that is “central” to its crafting of forecasts and policy yet has little grounding in real-world empirical data. Like active mangers who use the CAPM and DCF models, the Fed needs these bad models for planning purposes.