In mainstream economic models, consumer’s behaviour is generally assumed to follow a ‘utility function.‘ Consumers derive utility (creatively measured in ‘utils’) from whatever they consume, and they will attempt to maximise this subject to their budget constraint – and, perhaps, at a later level, some extra terms to incorporate behavioural quirks, social pressure or what have you. Unfortunately, even with modifications, the concept of utility is an explanation of behaviour that is questionable at best.
The first conundrum – as posed in the title of this post – is exactly what form utility takes. Is it supposed to be some sort of cumulative attribute that people collect as they go through life, like a stat on a video game? Or is it a temporary sensation experienced after consumption, so that economic agents are effectively utility junkies, chasing around temporary highs? There may be a case for regarding anyone who truly maximised utility as clinically insane and in need of help. In any case, thoughtlessly following predetermined utility functions leaves neoclassical agents with no real room for ‘choice’ – we know what their behaviour will be in advance, and it is unchangeable.
There is also the problem of fungibility: is it fair to suggest that joining a gym gives someone the same kind of satisfaction as eating a donut? Or that eating a donut gives the same feeling as owning a car? These nuances are lost in the aggregated world of ‘utils,’ a unit which has no relation to anything else and hence is hard to verify – at its worst, utility is simply circular: only measurable by the same behaviour it supposedly explains.
Economists have a standard response to contentions that utility is unrealistic. They will assert that, even though utility doesn’t really ‘exist’ – a position few would endorse, surely – it still follows that if preferences follow economist’s axioms, then an effective utility function can be derived. That is: utility is not meant to be taken literary, but economist’s assumptions are sufficient to ensure a relationship between preferences that is functionally the same thing. So it would appear the only way out for opponents of utility is to critique the axioms. I don’t believe this is true, but the axioms are worth critiquing before I explain why.
The two most important axioms required to derive a basic utility function are completeness and transitivity. There are other axioms that are also commonly used – independence, non-satiation, convexity – which are all vulnerable to criticisms, but since they pertain to the the exact form of a utility function, rather than the concept as a whole, I will focus only on completeness and transitivity. Without these, there is no utility function, whichever way you paint it.
The first axiom – completeness – is the idea that all relevant decisions can be definitively compared to one another: that is, there is no room for ‘I don’t know.’ There are clear problems with this. Often, it is hard to choose between two options, particularly if one is a bundle of many goods (e.g. two shopping baskets). In fact, as a decision rule this is generally computationally impossible. So people may act based on chance or impulse; they may seek advice or ask someone else to make the decision for them. What’s more, often people find it difficult to evaluate choices even after they’ve made them. Sometimes there is no ‘correct’ choice!
The other axiom, transitivity, implies that people will be consistent in their ordering of preferences. If I prefer A to B, and B to C, I will prefer A to C. It is an important axiom because, even if preferences are complete, a violation of transitivity means that utility functions can basically have any shape and therefore be pretty useless for clear calculations. While I expect numerous behavioural quirks suggest transitivity may be violated under certain circumstances, overall it is a fair axiom – for the individual. However, it has been known for some time that, once we have more than two agents, it becomes impossible to establish a clear, consistent ordering of preferences for the group. This isn’t moving the goalposts: it is highly relevant when we are using representative agents for the entire economy. (This problem also applies to the aforementioned independence axiom).
My most important point, though, is that even if preferences do follow all the axioms, utility is still highly flawed. This is because, like so many neoclassical models, all utility functions give us is a static snapshot of the economy (or individual) at a particular point in time, and there is no room for change. The simple fact that preferences are highly volatile and will be different in the morning and the evening, or in summer and winter, is enough to render utility useless for practical questions about the economy, which must surely incorporate time. Similarly, preference reversal has shown that the way options are presented has a large impact on the choice made by somebody, suggesting again that underlying ‘preferences’ are highly subject to change, and not really useful for the practical purpose of predicting behaviour. One can only wonder how utility might deal with a theory such as multiple selves, which would surely create the aforementioned aggregation problems for preferences, but for one person!
Now, I can almost hear the cries of “ah, but what is your alternative?” Actually, that doesn’t matter for the immediate critique. If I have a map of London Underground and I’m in New York, I’m not going to use it (even less so if I have a map of a fantasy land that exists only in the minds of economists). To push the analogy a little further, it is worth asking what I would do in this situation. I can think of two possibilities: either ask for help, or follow some simple rules of thumb based on what knowledge I have. This is the strategy economists should adopt.
In the case of ‘asking for help,’ what I mean is that economists should turn to other social sciences; namely, psychology, which has a far more empirically driven methodology than economics and has numerous explanations of behaviour. Economists truly interested in understanding human behaviour – rather than preserving their favoured assumptions – should collaborate with psychologists to create sound behavioural foundations.
Until then, economists should be content with simple empirically observed rules of thumb and intuitive aggregate relationships (they already do this with the marginal propensity to consume). Objections of ‘but Lucas Critique‘ are special pleading, since preferences are also liable to change with political decisions. In fact, I’d shout ‘Lucas Critique’ right back at economists, and suggest that they spend less time on the impossible task of making their models ‘immune’ to the Lucas Critique, while spending more evaluating the ever-changing relationship between policy and observation. It is better for economists to be vaguely right than precisely wrong.
Out of all the concepts in neoclassical economics, none is more imaginary, absurd and empirically falsified than utility. Economists supposedly follow a methodology of strict positivism, and based on the experimental evidence against utility, there is surely no reason to keep it. Yet for some reason, it doesn’t seem to attract the same level of criticism as other areas of neoclassical economics. Personally, I am puzzled as to why.