What is utility? Towards a model of human flourishing

Perhaps the most fundamental concept in economics is utility. Individuals seek to maximise their utility, firms, governments and other actors respond by providing the goods that give people utility, and the economy emerges as a natural consequence. Yet utility remains a rather under-investigated concept. There are good reasons for this. Having utility as a black box allows us to treat it as entirely subjective; this allows economics to retain its amoral character. However, it also encourages economists to focus almost entirely on GDP growth as the source of higher utility and this has come to be an increasingly untenable 

assumption of late.

Multiple streams of inquiry, notably those around the Easterlin paradox and the Genuine Progress Indicator, have noted that subjective wellbeing does not increase markedly above a certain level of income. Stated more formally, there are diminishing returns to utility from income. Graphically, this would look something like the following:

The black line represents the expansion path of utility as income rises. Note that it is concave. This is distinct from the neoclassical model where the expansion path is a 45 degree line i.e. linear. What's captured in this distinction is that as income (I) rises in the traditional model utility (U) rises proportionally into infinity. With a concave expansion path, as income rises utility rises too, but not proportionally. Indeed, as income approaches infinity the marginal benefit of additional income will approach zero. 

While the impact of GDP growth on utility from consumption is diminishing its impact on other factors is increasing. Environmental pressure from greater fossil fuel use, resource intensity and population growth is coming up against planetary boundaries. In the future, sources of GDP growth that derive from increases in economics throughput will be untenable. Unfortunately, historically GDP growth has tended to parallel increases in throughput. Moving to a system focused on increases in productivity rather than increases in GDP will be difficult and require a change in thinking.

Some scholars working on the so-called ‘Genuine Progress Indicator’ have suggested that what is required is an emphasis on development rather than growth. They propose that we need some way to measure human progress that is not predicated on increases in throughput. It is here that a richer understanding of utility becomes critical. If we can determine a way of increasing utility that do not revolve around the consumption of physical goods then we can ‘develop’ without growing.

It seems to me that if we want to build such a model we need to unite three separate streams of inquiry from economics, psychology and philosophy. Thus far these three have operated in siloed discursive spaces, to the detriment of each. Philosophy has largely ignored the importance of happiness and focused instead on meaning and virtue. Economics has focused, as already discussed, almost entirely on the physical determinants of life satisfaction. And psychology has devoted its efforts predominantly to happiness in the ecstatic (that is fleeting) sense rather than trying to model a ‘good life’ (though psychiatry has made some inroads at times—see Victor Frankel).

I propose a preliminary model of the following form:

Utilityi = Wellbeingi + Happinessi + Meaningi

The three variables—wellbeing, happiness and meaning—correspond approximately to the determinants each discipline has focussed on to date. That is to say, ‘wellbeing captures, for the most part, those determinants of utility that economics has traditionally focused on and which are improved by income growth. Things like food, water, shelter etc. ‘Happiness’ collects those determinants of utility traditionally investigated by psychology. And ‘meaning’ refers to those determinants traditionally investigated by philosophy.

This is a regression model, which I think is very important. Thus far we tended to focus on individual predictors of happiness in isolation. For example, we have studies (sometimes experimental for good identification) of things like achievement, companionship and money, but we don’t, as far as I know, have much in the way of regression models establishing the relative effect on happiness of all of these variables taken together. That is to say which don’t know which ones are most important or have the strongest effects. We don’t really know how they interact or whether their effects are linear, among other things. This means that if we were to guide public policy along these lines we wouldn’t know how to weight each factor, which would lead to enormous inefficiency. A regression estimate would correct, at least to some extent, this issue.

At this stage I have only very preliminary thoughts on what might be contained under each heading. Wellbeing can be approached formally through Maslow’s hierarchy of human needs and we can use traditional economic variables to measure it. Income is a good aggregate proxy because it enables higher quality food, water, clothing, shelter and even access to sex (though not reproduction or love, which are perhaps more important). The other two blocks—happiness and meaning—are not necessarily more difficult to approach formally but they are more difficult to find variables for. This is, I think, one of the strongest reasons why we have tended, in public policy at least, to focus on income growth as an indicator of progress—it is easy to measure.

I have written about happiness in a previous post. The core of the difficulty is that happiness is fleeting by definition, so we need time series models and high frequency data. It is no good discovering that heroin use leads to increased happiness if we can’t see that this happiness only lasts three hours and turns to unhappiness in the long run. It is also no good noting that silver medallists aren’t happy at the time they come second if we don’t observe that their training period was immensely purposeful and gave them a great deal of satisfaction, for example.

Meaning is not quite so hard. In the four thousand odd years of philosophical writings on utility we have some themes that can be approached with variables—love, family, religious connection, community engagements, among others. What is very difficult is capturing psychic health in the psychiatric sense. Someone with mental health problems of the chemical variety might be obvious in the data (there may even be an appropriate indicator variable) but someone who is merely highly neurotic (in the manner of Nietzsche, for example) will not show up. What is somewhat less difficult is measuring the determinants of meaning identified by the existentialist philosophers of the mid-twentieth century. Expressed in an extremely simplistic form, their philosophy is that life will be experienced as meaningful if an individual is able to satisfy their authentic preferences. What is meant by authentic is that the preferences are indeed the individual’s—they are not following the preferences of someone else, such as parents or the community, in bad faith and hoping for a good outcome nonetheless.

I have written about this particular problem in another post as well. What’s key and what has not been done satisfactorily in the past to my knowledge is the simultaneous investigation of preferences, life circumstances and utility outcomes. Utility has only ever been investigated against either preferences or life circumstances. But that is not what the theory from philosophy would suggest is important. A second, more technical, issue is that identifying preferences is tricky. It is especially tricky to identify complex preferences (such as for a career as an international trade lawyer rather than merely a career) and preferential trade-offs (such as for family or career) from data.  

So this is where I am at right now. My PhD will basically involve building up this model as far as I can in three years or so. Thoughts and comments from people greatly appreciated