members) might result in a decrease in subjective utility! This hints at what might be partial explanation for the effect described by Kahneman and Deaton: being less wealthy than average is itself a source of negative subjective utility.
This suggests that the relationship between wealth and utility also fails to satisfy condition (1) from Section 5.1.1. Given a group of people (neighbors, for instance), the differential equations describing the change in utility of members of the group relative to their changes in wealth will resist decomposition, because their utilities are a function not just of their own wealth, but of the wealth of other members of the community as well. By decomposing the system into component parts, we would miss this factor, which means that even if we took the principle of diminishing marginal utility into account in our calculations, the decompositionalist approach would still fail to capture the actual dynamics of the overall system. A more holistic approach is required.
This suggests an important lesson for the study of natural systems in which non-linearities play a significant role: the presence of unexpected feedback and variable degrees of mutual influence between different components of a system might well mean that attempts to model the system’s behavior by way of aggregating models of the components are, if not exactly doomed to failure, at least of very limited use. We must be extraordinarily careful when we attempt to tease general predictions about the future of the global climate out of families of EMICs for precisely this reason. We shall return to this point in Section 5.2, but first let us turn our attention to the other central challenge to be discussed here: chaotic behavior.