What's wrong with subjective wellbeing?

In happiness research we tend to build models around survey questions pertaining to ‘subjective well being’. They might take the following form: “Have a good long think about your life right now. Think not so much about this moment, but about perhaps the past year and the trajectory you are on going forward. On a scale of 1-10, how happy do you think you are?” These questions are very powerful for measuring aspects of human flourishing that might evade conventional economic indicators like life expectancy and income. But they have their own problems. Moreover (or perhaps more accurately, depending on your perspective), we should be trying to measure and analyse other things as well.

I have two main concerns with subjective wellbeing: you don’t really know what you’re measuring; and it’s not a variable that can continuously increase as life gets better. Let’s take each of these issues in turn.

What is subjective life satisfaction actually measuring? Is it ‘happiness’? That’s what a lot of commentators suggest and even seem to want it to measure. I don’t think this is correct, or desirable. Happiness is a slippery term, but if we want to meaningfully include it in a theory of human flourishing alongside concepts like meaning, affirmation etc, then happiness should (and does, I think) refer to a fleeting sense of ecstasy that accompanies accomplishment. You are happy when you score a goal in a football game, or get a promotion, or get laid. You are not happy all the time. To ask someone to think about a large section of their life and then ask whether they are ‘happy’ is not appropriate.

You want to ask people whether they are flourishing, which is distinct from happiness. Indeed, flourishing appears to require periods of unhappiness or at least adversity and struggle. You cannot get promoted if you are not working hard to earn that promotion. But presumably when you arrive back from work exhausted you are not ‘happy’, though you might be flourishing, experiencing your life as meaningful, contented etc.

Measuring happiness would be best achieved through some kind of time series study with high frequency data. You would have to monitor people for a significant amount of time, say a month at least, and ask them to register when they were feeling happy/sad/average. You could then associate changes in mood with extrinsic events. Additionally, you could measure how large a spike in happiness was associated with different events and perhaps how long those shocks took to die out.

The danger with basing public policy on ‘happiness’ is that you either try to eliminate suffering or keep people in a perpetual state of ecstasy, neither of which is very healthy or even , I would argue, what the human spirit needs. Keeping people in ecstasy is essentially what the drug Soma does in Brave New World. Every time people feel a little down, or doubtful, or confused or overwhelmed they pop a bit of Soma (which seems to be very similar to the aptly named real world drug ecstasy) and they’re right back up there. People in the Brave New World are ‘happy’ all the time, but they also lead incredibly vacuous lives. Is this a sensible thing for public policy to aim for?

Eliminating suffering might seem like a reasonable goal at first glance. Indeed, it is arguably what contemporary public policy aims for through its emphasis on GDP growth. GDP is a good proxy for health outcomes, education outcomes, incomes, consumption levels, comfort, access to luxury and entertainment etc. Suffering is, on the face of it, a bad thing, so eliminating it seems sensible. Moreover, as Karl Popper pointed out in his essays in favour of negative utilitarianism, as a society we can typically agree on what is bad and should be eliminated, but not on what is good and should be promoted. So focussing on suffering seems sensible.

So what’s the problem then? Well, as I already mentioned, a good life sometimes requires suffering. You can’t experience the glory of winning a marathon if you don’t hurt through hours of training. You are unlikely to ever get laid unless you have the guts to ask a few people out and get rejected a couple of times. Now obviously these aren’t the kinds of things public policy is looking. But when you move from approaching the elimination of suffering through the increase of GDP to instead eliminating suffering by looking at subjective well being, you get into trouble with this kind of stuff. In trying to eliminate suffering from people’s self reported wellbeing, you may end up eliminating all substantial endeavour from their lives as well. This may work for some, but not for most. Even the waitress who famously won the lottery went back to work at her minimum wage job within a week because she felt her life would be empty without it.

So subjective wellbeing doesn’t measure happiness because happiness is fleeting, and we wouldn’t want it to anyway because then we’d end up with misguided public policy. What else might it measure?

Perhaps it measures what it says it measures—subjective well-being: how someone is going, generally speaking.

The issue here is homeostasis. Simply put, human neuro-chemistry is evolutionarily designed to bring us back to an even keel, say 7/10 on the subjective well-being scale. I can’t say with certainty where the homeostasis point is on average, but initial samples I’ve looked at suggest around 7.1/10. One further concern in this regard is that homeostasis is likely at different points for different people. Some people’s baseline might be at 6/10 (sour types) while or others it might be at 8/10 (glass half-full types). This makes sense evolutionarily—we would want, as a species, some people to be a bit more pessimistic and others optimistic.

All this makes for quite a quandary when you’re making inferences from subjective wellbeing measurements. When someone answers 7/10, what does it even mean? Are they a sour person who is having a good year? Or a positive person having a bad one? Or just an average person having an average year? Moreover, what are we comparing their experience to? It is quite well known now that subjective well-being is roughly the same between underdeveloped nations in Eastern Europe and the highly developed nations of Australia and Scandinavia. We can make some inferences from small differences in averages, but homeostasis makes it difficult to talk about those small differences with anything approaching rigor. I have some hope for panel data models, but I suspect they will constantly be under-identified.  

I am currently thinking of ways to account for homeostasis. One idea I have is to use several control groups from within a population with some homogeneous features. For example, I could have three samples all drawn from Sydney: one from the Rich Northern suburbs, one from the middle-class Southern districts, and one from the outer-West working class areas. I could establish baseline subjective wellbeing levels, which would probably be similar for each group as they are all from Sydney, and then introduce similar shocks into each group, such as income increases, and see what happens. I don’t have a hypothesis at this point, but I imagine something interesting might emerge, especially if I use CPS data from the states, which goes back decades, or, if quality data exists, Japanese/Chinese data, which would show dramatic improvements in baselines and big spikes over time.

This point about homeostasis leads into another issue with subjective wellbeing: it doesn’t give you meaningful data on human flourishing. What is human flourishing? A big question but let me try a brief answer (I have posted longer ones in many places and will do so again shortly). It is a combination of three things: wellbeing, happiness and meaning. Wellbeing refers to physical health. It is typically assured by the traditional measures of economics development: income provides health care, shelter, food and clean water. If we wanted to be more expansive in our definition we might refer to Maslow’s hierarchy of needs. Happiness refers to that occasional feeling of ecstasy that I have mentioned. It is hard to go through life without any joy. But it is also insufficient to have only joy in life if that joy is amidst the ravages of cholera, famine and other hardships. Meaning is the tricky one. Without getting too philosophical, it refers to the need to have values and to affirm those values in the world. Note that values include ‘meanings’ and ethics. Someone who campaigns vehemently for civil rights is liable to have a high degree of meaning in their life, but someone who is retired and spends their time volunteering time at the local bowls club probably does as well. Most corporate careerists, who spend little time genuinely socialising or contributing to the community, may also have high meaning from the affirmation of their career aspirations.

Why does subjective wellbeing struggle with all this? Well on the one hand it doesn’t. This bundle of ideas: health, happiness and meaning, is quite nebulous, and nebulous notions like subjective wellbeing are useful for capturing the whole of the nebula. But that is also the weakness from a scientist’s point of view. What are you capturing and how can you draw inferences? Is this person you have just discovered is an 8/10 a generally happy person just coasting along, or are they are an unhappy person who leads a very meaningful existence? Is this 6/10 person leading a vacuous existence with few sources of joy but supported by exceptional determinants of physical health? Or are they someone who desperately wanted a certain promotion to feel fulfilled and failed to get it?

One idea this points to, I think, is that we need to have a good idea of a particular person before we can understand what they mean when they give certain subjective wellbeing scores. Often we can get this information directly from responses to other questions in the particular data set we are looking at. For example, the HILDA data set includes questions on how important community and religion are to a person. In other cases the information might need to be dredged out or assumed on the basis of available information. For example, HILDA might provide us with data on the importance of financial security to a person, but it will not give us information on how much income is required for that person to feel secure. We can, however, potentially infer this from a time series comparison of their income and life satisfaction scores.  At a push, we might also be able to dredge it out of an analysis of where they traded off hours of work for hours of leisure.

Let’s move now to discuss the second issue I mentioned at the start of this little brainstorm: subjective well-being is not a variable that can increase as life gets better. It is fundamentally bounded between 1 and 10. This makes sense for a variable subject to a stable mean (homeostasis), but it is very worrying if we want to make policy decisions based on subjective well being. Consider this: you are 22, you’ve just got a good grad job and a new apartment that looks great. You register 9/10 on the subjective well being question. Fast forward a bit and you’re 35, you’ve just been promoted to partner and you’ve married the woman of your dreams. You register 9/10 on the subjective well being scale. Life, it seems to me, has gotten better, but you’re still scoring the same.

Now arguably the inference from this story is that policy should focus on giving people the tools to score 8+ continuously from a young age. It should get them on a certain track that will inevitably lead to sustained high levels of subjective well-being. I don’t dispute that argument, but I don’t think the subjective well-being measure is very helpful with regards to it.

Consider this, for starters: what wellbeing score were you registering in the years between 22 and 35? How did you feel while doing your graduate job, or while you were senior associate working long hours? You were on a career track and meeting goals, but those goals were several years apart. You might have gone through an intimate partner or five in the meantime, with commensurate ups and downs in your mood. Perhaps your wellbeing dropped to 8 or even 7 over this period, but it is required for the 9/10 at the end. Public policy would seem to be failing over these down periods between spikes, but actually we are just suffering (or investing perhaps) on how way to a big payoff. A second point illustrated here is, very roughly, that if you enjoy life quite a lot when you are young then your even better life when you are older won’t really register on your subjective well-being scores. Even in time series there might only be small rumbles rather than the steady upward trajectory or spikes that a policy maker would find useful as far as insights were concerned.

I should emphasise a previous point again: historically, the story of this individual would have been redundant to a policy maker—they would not have been interested in people scoring 7+ consistently, only in those who were down in the dumps. But the move from GDP to subjective well being (or things like the GPI, which are similar) is predicated on focussing on what makes people flourish. The movement begins from the idea that GDP has run its course as a useful variable, and now we must focus on something else. If we are to use subjective well being in this case then we need to be interested in getting people to 9/10 on a regular basis.

Let’s dwell on this for a moment. If you’re an Australian policy maker you’re currently working with a population of people scoring well over 7/10 on average. How do you think about making their life better? We have plenty of suggestions: more money, more equality, more community, more public art and a better global environment, among other things. How large a change are these things likely to have if we introduced them into a population that is already so happy? Would it register? What kind of a magnitude of a shock would we need to get a rise out of the population?

Now obviously there are going to be ways to measure these things and make inferences, but I suggest that it will be more difficult than people sometimes think. It will be especially difficult to establish causality rather than simply correlation (and weak ones at that). We will have huge difficulty replacing our current paradigm of GDP increases with increases in ‘development’ if we use subjective well being as a proxy for development because GDP can rise infinitely while subjective well being cannot. We will also have problems using subjective well being as a measure for the success of our society because subjective well being is heavily determined by homeostasis. Anti-growth types will use high subjective wellbeing as a justification for anti-growth policies, but in doing so they will overlook the fact that subjective well being says little (potentially) about whether someone’s life is meaningful or not, rather than just pleasant, and we may well need more growth in order for everyone to be able to enjoy a meaningful rather than just pleasant existence.  

These are just some rough thoughts I’m having at the moment, but I intend to explore them further over the next few years as I embark on a PhD dedicated to these and related issues. If you have any comments or thoughts, please pass them on.

This article was first published on markfabian.blogspot.com on 15 January 2015.