Dynamic Stochastic General Equilibrium: macroeconomics at a dead end

The DSGE models are not only seriously flawed. They seem eminently irreparable. Their early versions started with over-simplistic assumptions and produced irrelevant conclusions. The later benchmark versions inherited the simplistic core on which ad hoc ‘frictions’ were grafted. These versions proved spectacularly useless. The most recent versions promise to make amends by allowing for agent heterogeneity and more attentive treatment of the financial aspects of micro-behaviour. But this promise cannot be squared with the basic ideas underlying the approach: general equilibrium and super-rational behaviour (including formation of expectations) attributed to so-called ‘representative individuals’. The derivation, specification and estimation of computable DSGE models involve a great deal of pure arbitrariness while their identification remains problematic. A science of micro-founded macroeconomics may be impossible. But scientific macroeconomics not based on micro-foundations may be possible. Newton’s laws of movements of solid bodies belongs to science without being derived from models of behaviour of elementary particles.