David Hendry on Why Forecasting Fails

The noted econometrician writes: “The intermittent failure of economic forecasts to ‘foresee’ the future reflects both imperfect knowledge and a non-stationary and evolving world that is far from ‘general equilibrium’ and closer to ‘general disequilibrium’.”

“[This has] disastrous consequences for dynamic stochastic general equilibrium (DSGE) systems, which transpire to be the least structural of all possible model forms as their derivation entails they are bound to ‘break down’ when the underlying distributions of economic variables shift. This serious problem is highlighted by Hendry and Muellbauer (2017) in their critique of the Bank of England quarterly econometric model (BEQEM–pronounced Beckem: […] a DSGE which, as in the film ‘Bend it Like Beckham’, bent in the Financial Crisis, but so much that it broke and had to be replaced.”

“Surprisingly, despite that abject failure, it was replaced by yet another DSGE (COMPASS: Central Organising Model for Projection Analysis and Scenario Simulation […]. Unfortunately, […] COMPASS had already failed to characterize data available before it was even developed. Persisting with such an approach introduces a triple whammy as:

a] the derivations sustaining DSGEs use an invalid mathematical basis;

b] imposing a so-called ‘equilibrium’ fails to take account of past shifts;

c] the DSGE approach assumes agents act in the same incorrect way as the modeller, so assumes agents have failed to learn that imperfect knowledge about location shifts forces revisions to their plans.”

“During a visit to LSE in 2009, Queen Elizabeth II asked Luis Garicano “why did no one see the credit crisis coming?” to which a part of his answer should have been that DSGE models dominated economic agencies and essentially ruled out such major financial crises by assuming away imperfect knowledge. Prakash Loungani (2001) argued “The record of failure to predict recessions is virtually unblemished.””

The article is available from the here.

Posted by at 10:33 AM

Labels: Forecasting Forum


Subscribe to: Posts