Modelling imprecise preferences and identifying the implications for theory and practice.

  • Loomes, Graham (Principal Investigator)

Project Details


If we want to understand how the economy behaves we need to understand how individuals make decisions: how they measure the desirability of different goods and services, how they compare benefits and costs, how they judge risks and weigh them alongside the possible good and bad outcomes, how they balance present and future considerations.

Traditionally, economists have tried to explain and predict behaviour on the basis of a model where human decision makers are assumed to operate as if according to some set of highly-articulated preferences which reflect their best interests, and where their actions constitute the rational pursuit of those interests. The same assumptions also underpin standard welfare economics and the policy guidance that flows from that analysis.

However, there is now an impressive - some would say, irresistible - accumulation of evidence suggesting that such a model is at best too simplistic and at worst misconceived and unreliable as a basis for prediction and prescription. There has been some momentum building during the last three decades to incorporate psychological insights and behavioural realism into models. But as yet we have not been able to accommodate more than a subset of the evidence within any particular variant or modification of the standard model.

This may be a sign that we need a more radical approach to modelling the problem. The proposal here is to start from the premise that people's underlying values and preferences are often only partly formed and imprecisely articulated (as opposed to the usual modelling strategy which is to start with some deterministic 'core' theory and add 'white noise' to mop up the factors not fully specified) and to examine the implications for decision making and behaviour of a model built on such 'shaky' foundations. A recent paper (Butler and Loomes, American Economic Review, 2007) shows that such models, even when only loosely specified, may actually have enormous potential for organising some of the most problematic violations of standard theory - and indeed, have the power to predict fresh violations not previously observed.

So the research will gather data through controlled experiments which will provide information we do not currently have about the variability of people's values and preferences and their degrees of confidence in their judgments. It will extend the investigation to include the ways in which people behave interactively to try to understand whether the imprecision-based model can account for some of the best-documented failures of standard models to predict patterns of strategic interaction. The potential importance of this is to help provide new models of, say, 'boundedly rational' consumers, and thereby better understand failures in the operation of markets and help improve the strategies that governments might adopt to repair or regulate markets more effectively.

The research will also examine another area of major practical policy significance: namely, how to collect reliable data from members of the public about their values and preferences for such things as health and safety improvements and environmental and social quality. Surveys designed on the basis of conventional assumptions have repeatedly generated patterns of behaviour that do not conform with those standard assumptions. As things stand, it is difficult to know how to interpret those results and how best to use them to guide public policy. However, a different modelling strategy may be able to give useful insights and enhance both the design of elicitation methods and the translation of results into robust policy guidelines.
Effective start/end date1/04/0931/03/12


  • Economic and Social Research Council: £440,535.00