Humans, not agents

In my early morning browsing, I read a paper called ‘Time to Abandon Group Thinking in Economics’ by Sergio Da Silva (pdf). It isn’t as clear as it might be, but the line of argument concerns the unscientific status of the representative agent approach to macroeconomics. In conventional macro, he writes: “The whole is viewed as merely the sum of the parts. Of course, this would be so if the constituent individuals were homogeneous. But they are not. Despite that, macroeconomics assumes homogeneity of individuals and focuses on a “representative individual.” Rather than explaining collective behavior from the interactions between the constituent individuals, macroeconomics studies the behavior of the average individual.”

This is not justified, he continues, because the macroeconomy does not empirically exhibit the property of “self-averaging,” such that as more and more individuals were aggregated, a central limit theorem holds – or, in a Poisson distribution, the model coefficient of variation approaches zero. To put it another way, what benefits an individual need not always benefit the group – there are losers as well as gainers.

The paper goes on to advocate applying the aggregation tools of statistical physics and biology to macroeconomics. This was one of the alternatives discussed at an international macro symposium the ESRC organised in Oxford last October, presented by J.P.Bouchaud & summed up in his essayEconomics Needs a Scientific Revolution. I think almost anything would be better than conventional macro, addicted to representative agent DSGE models despite their non-compatibility with any evidence. However, there’s no consensus (not surprisingly) about the alternative, and three were presented at the ESRC symposium. The other two were network approaches and complexity approaches.

The lesson macroeconomists should take – the point of this ramble – is that the social sciences need to be consistent with the biological and human sciences. This is essential for economics to move from being “applied logic” as Da Silva describes it in the paper to an empirical science.

There are two domains of knowledge to be incorporated. One is cognitive science and neuroscience, and there has been some progress here in microeconomics, with behavioural economics and neuroeconomics. Last year I attended a fascinating workshop at the Toulouse School of Economics that asked what cognitive science could tell economists about attention and therefore economic decisions – it resulted in my conference report The Invisible Hand Meets the Invisible Gorilla (pdf).

The other is evolutionary biology and ecology –  and macroeconomists display no interest in what these domains can teach us about aggregation and group behaviour. Business economists have long used evolutionary metaphors in an intuitive way, and evolutionary economists such as Geoffrey Hodgson in, for example, 

have tried to formalise models of behaviour at the level of markets. But – as I found when writing about these areas for 
– there has been scant mainstream interest.

[amazon_image id=”0472084232″ link=”true” target=”_blank” size=”medium” ]Economics and Evolution: Bringing Life Back into Economics (Economics, Cognition, & Society)[/amazon_image]

Da Silva’s paper cites plenty of natural scientists but next to no economists. This must be a mistake by our profession.


7 thoughts on “Humans, not agents

  1. Pingback: A week of links - Evolving Economics

  2. Pingback: Does economics need *more* physics envy? | The Enlightened Economist

Comments are closed.