Illustration: Jayachandra/Mint
Illustration: Jayachandra/Mint

An economics to fit the facts?

Economists must rely more on a nuanced understanding of human behaviour while rendering policy advice

The discipline of economics is by consensus one of the major casualties of the recent global financial crisis. Fiscal multipliers and monetary policy have not performed as modelled, the much-programmed financial markets have not worked rationally, near term economic forecasts based on the International Monetary Fund (IMF)’s intimidating dynamic stochastic general equilibrium (DSGE) simulations have repeatedly failed, and there continues to be great confusion even over something as basic as measuring growth potential. Barry Eichengreen recently pointed to some drawbacks in the economics discipline currently practised (

However, his empiricist pursuit of the holy grail of the perfect model through a new “evidence based economics" that will fit all the facts is likely to meet the same fate as all those that preceded it, such as the IMF’s new “MAPMOD" that embeds the financial sector in the tried and failed DSGE model. Historians have been there and moved on. It is time that the younger discipline of economics did likewise.

Latter-day economists, it seems, have still to reconcile themselves to the humbling fact that their discipline is a social and not a natural science; that their subject matter consists of human behaviour for which there can never be any natural laws beyond some universal foibles, such as greed, jealousy, selfishness, thrift, pride, prejudice, etc. personified in the classics of literature. These classics remain evergreen because people have related to them across space and time as they distil the essence of what it is to be human. That is why we can relate more to Thucydides’ History of the Peloponnesian War, which ascribed human actions to such foibles, than to Herodotus’ more famous History, which ascribed it to divine intervention. This is why we still relate to Micawber’s advice to David Copperfield in the Charles Dickens classic: “Annual income twenty pounds, annual expenditure nineteen six, result happiness. Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery."

This iron law of the consequences of leverage can scarcely be bettered. But it is too general a tool for model building. Not all humans are profligate. This is why all model building and universal laws relating to the human condition have floundered, most famously those devised by Karl Marx. It could plausibly be argued that while individual behaviour can vary because of personality differences, macroeconomics is about aggregate behaviour that is more predictable and therefore programmable.

To take an example from physics, while particle behaviour at the sub-atomic level is unpredictable, in the aggregate the physical world functions according to natural, immutable laws. Might not the same apply to human behaviour? Is human behaviour in the aggregate rational and profit maximizing as Adam Smith assumed, or subject to some subconscious, almost Freudian urges as recently underscored by behavioural economists such as Richard Thaler and Cass Sunstein?

There are, unfortunately, critical behavioural differences in response to similar circumstances across both space and time at the aggregate level. An increase in income can induce people in one part of the world to save more (think Angela Merkel’s Swabian housewife), others to spend more (think the stereotypical credit card-happy American housewife). Likewise, Asian societies behave like ants and save more than Latin American societies that behave more like grasshoppers. The response in the same place may be quite different in different periods, such as in pre-capitalist peasant societies that valued leisure more and cut back on production in response to higher prices.

Even to this day the French and Italians value leisure more than Americans and this is reflected in their economic behaviour. The same people behave quite differently in an economic boom and in a recession. Think “Ricardian equivalence", “Minsky moment" and Keynes’s “animal spirits". Data-dependent relationships, and even broader generalizations, derived from a specific time or place may not hold elsewhere or at other times. To get around this problem, historians turned to the narrative as an analytical tool, because describing the way things happened turned out to be not very different from explaining why they happened. We need to recognize that similar facts need not result in similar outcomes across space and time.

In these circumstances, is it surprising that IMF’s models do not yield accurate forecasts of growth even in the next quarter, except in times of exceptional macroeconomic stability where future trends are a replica of past trends? In financial markets, where sophisticated model-building drives investment, interconnectedness and the critical role that confidence plays can on their own gut any model, and they repeatedly have. The relationships between variables based on past data used to generate coefficients may not work going forward because there may not be any causal relationship between correlated data, and the correlation itself may not hold going forward. Models based on data derived from one area may not work in other areas.

The bottom line is that models and algorithms are useful, indeed desirable to avoid human error, where outcomes are dependent on physical and natural laws, such as flying aircraft or driving cars. The problem arises when algorithms are pushed to apply to areas such as economics and finance, where outcomes are contingent on human behaviour. Economists consequently need to rely less on complex universally applicable models and incorporate a more nuanced understanding of human behaviour as they go about reinventing their discipline and rendering policy advice.

Alok Sheel is a civil servant. These are his personal views.

Comments are welcome at