« "Obama's First Choice" | Main | links for 2009-01-21 »

Tuesday, January 20, 2009

"Models of Bounded Rationality and the Credit Environment"

When I talk about the need for government intervention in the marketplace, the justification for the intervention usually relies upon the presence of substantial market failures of one sort or another. However, as Robert Waldmann often points out in comments to those posts, government intervention can be justified in other ways besides the traditional list of market failures, e.g. the presence of dynamically inconsistent preferences can be used to justify forced saving for retirement. The discussion below looks at how governments ought to respond to the current problems in the economy if a key assumption of economic models, rationality, is dropped and replaced with an assumption that agents have bounded rationality:

Models of bounded rationality and the credit environment, by Leigh Caldwell, voxeu.org: Responses to the recession should not be based on unrealistic expectations of rational behaviour. We now know enough about real, flawed human psychology to be able to take some account of it in policy setting.

The “homo economicus” model of rational agents, acting to maximise utility in the possession of all available information, is not realistic. It is hardly a credible way to look at human beings – but we tolerate it because it is simple enough to allow equilibrium analysis which often gives reasonable predictions.

However, these equilibrium models are not serving very well in today’s situation. Standard monetary policy and (to a lesser extent) Keynesian theories are based on rational-actor assumptions. They give broad recommendations about monetary loosening and fiscal expansion – which central banks and governments are rightly trying out. But growth is not resuming.

Bounded rationality is the broad term for behavioural models that do not follow the rational-maximiser formula. There is not yet a generally accepted alternative model. Lots of individual non-rational behaviours have been discovered, but they are grafted onto a rather clunky ‘rational actor with bits’ instead of forming a coherent behavioural model.

However, the best place to test a new theory is often at the edges of the old one, where the existing model breaks down. So the current troubles in the financial and real economy may be a good opportunity to try out some alternative models and see which give a reasonable description of what we see.

Models of bounded rationality

Different models of bounded rationality vary basic assumptions of the rational agent model in different ways. Some of those assumptions are:

  • Utility is discounted over time in a consistent way
  • People have access to all relevant information
  • All relevant information is expressed through market prices
  • People can instantly weigh up the change in utility given by any buying or selling decision
  • People act to maximise their utility

Therefore, the typical way to create a bounded rationality model is to relax one or more of these criteria.

The first class of model explores different time discount rates. Experiments show that people apply a high discount on utility in the present – they prefer £100 now to £120 next year – but a lower one in the future – they prefer £120 in four years to £100 in three (Loewenstein and Prelec 1992). This, arguably, contributes to explaining recent huge variances between overnight and three-month interest rates.

Perhaps our psychological instincts reflect a pragmatic sense of the risk of a promise not being fulfilled, and the modern financial system has evolved to provide a confidence in the future that does not come naturally. If so, then the financial markets are no longer successfully filling that role. This short-long imbalance is an important cause of disappearing credit.

A second kind of bounded rationality reflects the idea that people do not have access to all relevant information. Danielsson and de Vries (2008) discuss an important example – information asymmetry – and an interesting solution – enforced transparency. Many agents, short of crucial information, have had to make guesses; clearly this has led to many of the write-offs we have seen in the last 15 months.

A third approach relaxes the constraint that prices express all the information needed to make decisions. This classical assumption works only when there is a sufficiently liquid market, with defensible property rights. If enforceability is at risk (for instance when there is a significant chance of bankruptcy) then we can use derivatives or insurance to provide secondary price information about the risk of default. If the market for these derivatives is not big enough or does not trade publicly, we lose that price signal.

In these situations, we need to use non-price signals to make economic decisions, and we do not have good models for interpreting those messages. This leads to inconsistent or irrational decisions on resource allocation.

A fourth category of model removes the assumption that people can make an accurate calculation of the utility that a decision will provide. These models include satisficing (Simon 1956) and rational ignorance (Downs 1957). Another is the concept of anchoring or habit – which suggests that we rely on familiar choices to avoid the mental effort and risk of picking alternatives. This behaviour can be seen in the credit markets, where an aggressive conservatism appears to be in control of lending decisions.

The fifth alternative to the rational model is to ask what happens if people do not act to maximise utility. Indeed, is there any single quantity called utility? Alternative models of decision making include multi-dimensional (“vector”) utility functions (for example Rios 1987), value modelling (Gordijn 2001) and psychological attempts to understand subconscious mental drivers – which have rarely been explored in economics. My own research suggests that a concept called “local utility gradient” drives behaviour.[1]

Intuitively, the current financial environment seems to expose a lack of rationality in market participants. The task for researchers is to find whether specific irrationalities provide evidence to support any particular behavioural models.

Policy implications

Whichever model of behaviour is assumed, policymakers should ask what the models indicate for macroeconomic policy. This question has been considered in a conference on “behavioural macroeconomics” at the Boston Federal Reserve in 2007.[2] No doubt the old solutions will work, given enough time. But today, with a much better understanding of economic behaviour, we could design more sophisticated solutions which would work faster.

A clear example comes from the concept of anchoring. Anchoring creates a tendency to fixate on one option too long when we might profitably switch to another – it limits the effects of all kinds of quantitative changes in policy. If a bank is scared to lend at a spread of 2.5%, anchoring implies it is unlikely to start lending when the spread increases to 3% or even 4%. This demands a qualitative change in conditions so that the bank can no longer make a simple marginal comparison but is forced to re-evaluate its likely returns from scratch.

The same effect is at play with fiscal easing. When state borrowing breached the government’s self-imposed 40% debt ceiling, the anchor was broken and little credibility may be given to promises of future prudence as debt reaches 45%, then 55% of GDP and beyond. However if a new anchor is created, say at the Eurozone’s 60% level, this creates a qualitative limit – immediately more credible than an arbitrary quantity. This anchor in turn can change behaviour in the government debt market because it gives agents a new focus for their decisions.

If the utility gradient model is correct, it is better to enter the recession sharply – even at the cost of a big immediate reduction in GDP – if it creates an expectation that we have hit the bottom. When people believe it can only get better from here, they will act accordingly – investment and spending will start.

I do not suggest that behavioural and psychological considerations are the only cause of the crisis. But they make a substantial contribution to its persistence.

The paradoxical conclusion is that it may not matter what new institutions or new rules are designed – as long as something is done. Any new framework gives agents a reason to abandon their anchor to fear; and gives them a chance to reattach themselves to hope. This – the converse of Keynes’ paradox of thrift – is what will ultimately rescue the world economy.


Danielsson, Jon and Casper de Vries, “Money Market on Strike”, Financial Times Economists’ Forum 9 November 2008
Downs, Anthony. An Economic Theory of Democracy. New York: Harper, 1957.
Gordijn, Jaap and Hans Akkermans. “A Conceptual Value Modeling Approach for e-Business Development”. Proceedings of the Workshop Knowledge in e-Business, pp 29-36 (2001)
Loewenstein, G. and D. Prelec, “Anomalies in Intertemporal Choice: Evidence and an Interpretation”, Quarterly Journal of Economics 107, no. 2 (May 1992), pp 573-597
Rios, S., “An Interactive Sequential Approach to Multicriteria Decision Making”, Extracta Mathematicae 2, no. 1, pp 26-28 (1987)
Simon, Herbert. “Rational choice and the structure of the environment”. Psychological Review, 63, pp 129-138 (1956).


1 This research is underway and not yet published, but will be published as “Multidimensional utility and a model of bounded rationality”, by Intellectual Business in 2009. A summary of this and other ongoing research is available at http://www.intellectualbusiness.org/
2 Janet Yellen’s speech to this conference provides an excellent summary of relevant literature.

    Posted by on Tuesday, January 20, 2009 at 09:23 PM in Economics, Financial System, Market Failure | Permalink  Comments (19)


    Feed You can follow this conversation by subscribing to the comment feed for this post.