Changes in individual and/or aggregate?

Abstract

The majority of papers reporting the results of experimental tests on individual behaviour concentrate on the average or aggregate behaviour of the subjects, even though different subjects may exhibit different kinds of behaviour. At the same time, it is well known that subjects are noisy in their responses and thus may change their behaviour during an experiment. The question then arises as to whether this noise causes average behaviour to change during an experiment. We take the data from an experiment on behaviour under ambiguity, which try to identify which preference functional subjects are using, and analyse it in two halves. We find that, while some individual preferences seem to change from the first to the second half, on average, preferences are unchanged (though there is a reduction in the noise exhibited by subjects in the second half). This is a reassuring message for those who want to draw inferences from experiments.