Veteran economists weigh in on ‘rebuilding’ macro theory

Several top economists have come together to give their thoughts on how it might be possible to rescue the core model of New Keynesian macroeconomics from its critics.

The series of papers, published in the Oxford Review of Economic Policy on January 5, tackle the question of how best to repair the dynamic stochastic general equilibrium (DSGE) model that underpins so much macroeconomic analysis. The articles are free to read until February 7.

Broadly, most of those contributing say the model should be kept more or less intact, but that it should evolve in several directions. Summarising the ideas, David Vines and Samuel Wills say four main changes are needed: emphasising financial frictions; limiting the use of rational expectations; introducing heterogeneous agents; and devising better microfoundations.

Some of the economists are more sceptical than others. Nobel prize-winner Joseph Stiglitz argues DSGE modellers have not used the right microfoundations, particularly in failing to incorporate the insights of information and behavioural economics. He prefers to move towards a completely different set of models, that draw on the insights of partial – not general – equilibrium, such as banking sector models.

Simon Wren-Lewis likewise challenges the DSGE model’s microfoundations, arguing that if DSGE had not replaced the more traditional structural economic models a few decades ago, the links between the financial sector and real economy might have been better explored before 2008.

In doing so, he defends a recent position advanced by Olivier Blanchard, that economists should use DSGE models for studying theoretical questions that permit a simplified version of reality, and different, data-based models for setting policy. Blanchard elaborates the point in a paper contained within this latest collection.

Wren-Lewis thinks Blanchard may not have been ambitious enough. “I want to go substantially further, and argue that the failure of academic macroeconomics to adopt an eclectic methodological position was the key reason why it was so unprepared for the global financial crisis,” he writes. Though he does not want to scrap DSGE models, their hegemony must be broken, he says.

Such an idea runs counter to the teachings of Robert Lucas’ famous critique, which suggests data-based models are inappropriate for studying policy questions. If you change a policy, old trends will cease to hold, Lucas argues. DSGE models avoid the Lucas critique by creating a system whose underlying rules are unchanged by policy interventions.

Wren-Lewis and Blanchard think absolutism on Lucas’ point is misplaced. Wren-Lewis writes: “At any particular point in time there is a trade-off between data coherence and theoretical coherence, and insisting that all academically respectable modelling should be at one extreme of this trade-off is a major mistake.”

Science has had to deal with the trade-off between bigger, more feature-packed models and more abstract, but easier to digest, models for a long time

Andy Haldane and Arthur Turrell

In this spirit, Bank of England chief economist Andy Haldane’s paper, co-authored with Arthur Turrell, a physicist by training, proposes agent-based modelling (ABM) as a basis for realistic testing of theoretical propositions. The paper was also published as a BoE staff working paper in November.

“Science has had to deal with the trade-off between bigger, more feature-packed models and more abstract, but easier to digest, models for a long time,” Haldane and Turrell say. Physics, for example, uses complex models such as ABMs to explore hypotheses or “discover new phenomena”. Once an effect is identified, it can be built into a simpler theoretical model.

Among the papers, Jesper Lindé’s goes perhaps furthest towards defending DSGE modelling and suggesting how best they could be adapted for continued use in a policy setting. The Sveriges Riksbank head of research argues they are “very suitable” for policy analysis, and an adapted version is likely to continue as a “workhorse tool” for a long time to come.

Lindé says that despite their failure to see the global financial crisis coming, DSGE models are not wildly inaccurate when used to study longer term trends, and did not perform much worse than other kinds of models in 2008. He also notes they proved useful in designing a policy response to the crisis.

Lindé suggests a useful way to proceed would be to ensure DSGE models are “modular”, and as such can incorporate the insights of smaller “satellite” models used to test specific hypotheses. He points out that without thinking in general equilibrium terms, partial models may fail when applied in a whole economy, hence the need for a large core model.

Like many of the other economists, he also concedes the importance of allowing several different modelling approaches to develop, giving particular praise to Haldane and Turrell’s ideas.