History of happiness studies and what I'd like to contribute

This is my previous two PhD presentations combined into one file for anyone who is interested. More material is available over at my blogs specifically for this stuff: modellinghappiness.blogspot.com

This week I’m going to talk about the history of happiness as a concept in public policy. As I’m doing a PhD in economics this will be mostly about economic attitudes to happiness, but as this is a public policy school I will discuss the concept more generally as well.

I think the best place to start would be the so-called Paretian turn, which is when economics split from psychology. What I’m about to say is based mostly of the appendix in Measuring Happiness by Weimann, Knabe and Schob; and The Road Not Taken: two debates about the role of psychology in economics by Sugden and Bruni.

At the inception of economics around the turn of the 19th century the discipline was very much related to psychology. In particular, the idea that humans try to maximise utility were derived from psychological results that suggested that humans approach pleasure and avoid pain, and the notion of diminishing marginal returns was derived from early work into desensitisation and related phenomena.

Pareto didn’t like this stuff for two reasons: first, it didn’t leave much space for economics as a standalone discipline, and second, it was quite messy because it depended on measuring cardinal utility, which was impossible.

Pareto advocated instead for an ordinal approach to utility and developed the notion of rational choice. Various assumptions could be made about the ability of individual’s to rank their preferences and then these preference order rankings could be used to derive mathematical expressions that could then be manipulated to prove various aspects of rational choice. Economics thus became the science of rational choice.

It strikes me as weird to call this a science because there is nothing empirical about it, but let’s leave that to one side.

Central to a lot of classic work in rational choice theory was the idea that there existed a strong relationship between income and utility. Indeed, as we would all know from microeconomics you can derive the expenditure function from the utility function. There is perfect coherence.

Economics kept chugging along with these premises for many decades doing great things until in the late 70s Richard Easterlin took the World Values Surveys of the Gallup organisation – some of the first data sets to ever include questions about happiness – and used those to empirically verify the assertion of a direct relationship between utility and income.

He made the strong assumption that happiness and utility are basically the same thing and derived the following principle results. First, there are diminishing marginal returns to happiness from income. Second, income graphically explains the differences in happiness between people very well within countries, but seems to explain very little of the result across countries. This is the so-called Easterlin Paradox.

The main explanation he presented for this phenomenon was reference points and relative status. If you’re a young Londoner, economics would say that you have much higher levels of happiness than a goat herder in rural Pakistan. But because you compare yourself to David Cameron, Kylie Minogue and your mate who got into Cambridge you feel bad, while the goat herder compares himself to his neighbour who has no goats at all and feels great. This explains the role of happiness within country and between countries.

In his first paper from 1974 (actually a book chapter) ‘Does Economic Growth Improve the Human Lot?’ Easterlin did not explicitly mention adaptation, but the seeds of this idea are present in his discussion of diminishing returns to income.

Psychology through the 80s and 90s would in turn provide some powerful insights into these phenomena: reference groups, relative status and adaptation. This period marks the beginning of a re-convergence between economics and psychology on the fringes of economics.

The most important book in this regard is perhaps Wellbeing: the foundations of Hedonic Psychology. One of the editors is Daniel Kahneman, who some of you may recognise as one of the founders of behavioural economics, which attempts to ground economics more thoroughly in empirically verified facts of human psychology rather than the rational choice theories neoclassical economics uses.

A few important ideas are summarised well in that book. The first is that adaptation is a very real phenomenon on a range of levels. We adapt to pleasure and pain and our memories of such phenomena are heavily conditioned by the peak intensity of feeling and the last feeling we had. So for example, someone who has a long colonoscopy but it is mild at the end will remember it as less unpleasant than someone who has a relatively brief colonoscopy that ends with a spike in pain.

A second key finding is that reference points exist for all kinds of things, not just status. That ‘peak pain’ I just mentioned is a reference point. Similarly phenomena occur with paraplegics and the like. After the accident, the reference point is still from when you could walk. But after a few years the reference point is now when you were incompetent with your wheelchair.

These two notions coalesce into the idea of a hedonic treadmill—you are always running to stay ahead of your adaptation (or trying to accelerate your adaptation in the case of negative things).

A third key idea is the notion of subjective wellbeing, which is not yet tightly defined at this point in the research but leans towards emotional stability and positive affect rather than life satisfaction. Recall that Kahneman and colleagues are mostly interested in measuring how pleasure and pain motivate behaviour, so that are interested mostly in short term phenomena.

By the time Wellbeing was published a bunch of cross-sectional studies of happiness had been published and some stylised facts derived. Married people, especially men, are happier. Religion makes people happier. Unemployment is really damaging to happiness. There are others.

Wellbeing was published in 1999. Where are we now?

Psychology research into adaptation has converged around the concept of a set-point. We have a stable core of affect – we stay around 7/10 and converge back to it. Our set-points are largely genetically determined: we are glass half-full or half-empty types, extroverts or introverts etc.

It is important to note though that set points relate to affect not to life satisfaction. Mood in particular is homeostatically protected. There are, however, some events that we don’t seem to entirely adapt to, like living in a warzone or the death of a spouse. This is because these events either influence life satisfaction rather than affect or because they are related to a fundamental existential threat. Both affect and life satisfactions are destroyed by starvation, for example.

Economics has dug deep into reference group and relative status effects. The most up to date study of this is Happiness and Economic Growth. It was released last year and uses data from China to study the effects of income growth and reference groups, largely through internal migrants, who are abundant and easily observable in China.

It is argued by some contributors, notably Easterlin, that income growth isn’t actually that important, especially because people don’t like the uncertainty that comes with structural change and because of the unemployment that coincided with the deep liberalisation of the Jeibao & Jeimin era.

Martin Ravallion critises this view for not taking the possibility of re-scaling into consideration. I’ll quote him:

“Economic development is a process of structural change, which changes people’s reference groups and scales. It changes how you think of the world where you live when you move from a village, where the reference group is very narrow, to a city with a very vast set of people at different levels of living. In that process, the scale of subjective well-being that we use is surely going to change.” (pg. 246)

Re-scaling is a very neat explanation of the Easterlin paradox—people have different scales for what a 1 and a 10 are, but it is very hard to measure.

Some dutch theorists, notably Bernard Van Praag, have done a lot of experimental work trying to determine the value of absolute and relative income to people. They find, on average, that happiness derived from income is comprised some 60% of absolute and 40% from relative income, but that this ratio shifts over as you get richer, so that for people in the top income decile a 10% change in income rank is worth a 90% change in absolute income (I’m paraphrasing here and the numbers are probably quite different).

Finally, in a very influential paper, Stevenson and Wolfers combined a whole lot of different data sets on happiness and found that there was actually a very strong and consistent linear relationship between log(income) and happiness both within and across countries. They explain the Easterlin hypothesis as an artefact of bad data.

One thing that I have not been able to discuss is the history and perspective of the psychiatric and population mental health disciplines on happiness. These branches of psychology have a lot of things to say about meaning and life satisfaction that I think would really benefit the other literature. I will discuss them at a later date.    

 What contribution I would like to make to the literature

Last week I discussed the history of happiness research, especially since the Easterlin Paradox in economics in the mid-1970s and the advent of Hedonic Psychology in the 1990s. In the past I’ve presented the model that I’m going to be building. This week, on Bruce’s suggestion, I’m going to talk about how my model fits into the literature. That is to say, I’m going to explain what contribution I would like to make.

The main thing I want to do is integrate what we already know into a comprehensive theoretical model that we can then use to guide future research.

Happiness research has been driven by empirics rather than theorising. This is partially because both economics and psychology have science-envy. Well actually, it’s because they want to be more rigorous than the humanities and more traditional social sciences.

Unfortunately, this has meant that the ‘theory of happiness’ has historically been a collection of stylised facts rather than a model that we build, test and refine. Consequently, there is a great of deal of confusion in the literature. The most obvious example is the interchangeable use of a large number of terms to describe the object of research: wellbeing, life satisfaction, happiness, human flourishing, utility, the good life, eudemonia etc.

I think it would be valuable to step back from the empirics for a moment to develop a better theoretical model of happiness than combines and can explain the stylised facts we have discovered.

This would help clarify some the confusion I just mentioned.

It would also get this particular research area back to good scientific method. In science we ideally want to have a good hypothesis that we then subject to a range of tests. If it passes these tests we can treat it as fact until it is refuted by a later test and/or a better hypothesis.

If we don’t have a good theory then when our hypothesis passes or fails tests we can’t be very sure about what exactly has gone right or wrong.

Let me give you one example. It is undeniable that when we ask people the standard happiness question “all things considered how happy would you say you are 1–10” captures psychological wellbeing (existential issues), subjective wellbeing (emotional issues) and life satisfaction (life course and economic issues).

As a result, when we ask a successful, neurotic woman who has just lost a small bet on a sunny day this question and she answers 7/10, we can’t infer much from a comparing this to a poor, extroverted man who has just handed in his PhD on a rainy day who also answers 7/10. We really don’t know whether it is someone’s subjective wellbeing, psychological wellbeing, life satisfaction or mood that we are measuring, or what their relative contributions are to whatever overall faculty we are measuring with this question.  

Consider the consequences. In a recent paper, Easterlin essentially suggested that people were happier in the immediate aftermath of the Tiananmen incidents in China than throughout the next 20 years of liberalisation because unemployment became more common. He makes this claim based on surveys using the question I just mentioned. I suspect he is wrong because he the period after Tiananmen saw a great deal of liberalisation in China, and the ten years after that saw little major change but still huge increases in income. I suspect, moreover, that he is ignoring the possibility of rescaling, but we can’t know until we get a better theory that allows us to take more accurate and specific measurements.

This brings me full circle. We need a better theory with which to make sense of our existing empirical measurements, and this theory in turn will require empirics to be validated. It’s an iterative process, but thus far I don’t think we’ve given theory its due.

A valuable theoretical model at this time would be able to explain the stylised facts that we currently have about happiness. It would also be able to explain them in a mechanistic way as this makes testing and drawing inferences easier.

What do I mean? For example, if we hypothesise that happiness and log(income) have a linear relationship with a slope of 1/3, then we have a mechanistic theory that is both quite detailed and very easy to test. I can also draw inferences about how happy someone on $150 000 ought to be.

On the other hand, if I say that happiness that happiness comes from virtue, then I don’t have a mechanistic feature (what is the numerical relationship between happiness and virtue?) and I don’t have an obvious criteria upon which to base tests.

For these reasons I will try to develop a theory that is parsimonious, mechanistic and can explain the stylised facts we have.

Some of these facts include the Easterlin paradox, relative status effects, the log relationship between income and happiness, the importance of self-actualisation for some and comfort for others, emotional volatility in otherwise very satisfied people, emotional happiness among bonded labourers and very high levels of life satisfaction among the very poor in developed countries.

My attempts to explain all the stylised facts have already led me to what I think is another contribution of my paper to the literature, which is the integration of insights from philosophy and the clinical arms of psychology into thinking about happiness.

This integration is important to explain the difference between psychological wellbeing as in mental health, subjective wellbeing as in emotional stability and positive affect, and life or existential satisfaction. I hope that one of the main contributions of my paper will be to make the distinctions between these things clearer and their contribution to human flourishing more salient.