GEORGE MACKERRON (inventor of Mappiness): Well, Alex has stolen the content on my first couple of slides. So I guess I’m George, and I’m going to talk to you about the economics of happiness and some of the findings of my own research.
As Alex says, I’ve got several hats, academic economist, creator of a study called Mappiness, which is largely what I’ll talk to you about today, which I think is the biggest momentary happiness study that’s been done, and also, as Alex says, co-founder of a startup. So I research election environment economics and behavioral economics and happiness economics. I also teach PhD students regular expressions, but no one knows that they need to know regular expressions until they come to my session. They don’t come to my session because they don’t know they need to know regular expressions, despite the fact that I advertise it with an XKCD cartoon. Anyway, I know something about measuring and possibly something about improving happiness.
And as Alex says, I do a certain amount of teaching in the course of all of this. So today I want to give you a very brief history of happiness economics, persuade you that happiness is worth measuring, and that happiness measurements are valid and useful. And the poster for this talk promises to answer this question: How happy would it make you be to be drinking by an estuary in the vicinity of riots the day after your football team lost unexpectedly at a full moon when Donald Trump had just won an election? So I’ll do my best to answer that question as we go along. And I can tell you right away that two of those factors aren’t important.
But I’ll leave you guessing for now as to which two those are. So a very brief history of happiness economics.
Now, around 100 years after Bentham, we have Edgeworth. He was the founding editor of “The Economic Journal,” and laid lots of the foundations of sort of modern utility theory. And Edgeworth went as far as to imagine something that he called a hedonometer, an idealized happiness measuring machine. And he waxed really lyrical about this and says “let there be granted to the science of pleasure what is granted to the science of energy. Imagine an ideally perfect instrument, a psychophysical machine, continually registering the heights of pleasure experienced by an individual, exactly according to the verdict of consciousness or rather diverging there from, according to a law of errors. From moment to moment, the hedonometer varies, the delicate index now flickering with the flutter of the passions, now steadied by intellectual activity, low sunk whole hours in the neighborhood of zero, will momentarily spring up towards infinity.” But unfortunately, Edgeworth’s peers were mostly not on board with this idea. Economists then, in the early 20th century, increasingly wanted to be thought of as scientists. They had a kind of physics envy.
And they certainly did not want to be thought of as psychologists such as Freud, who we see here. And therefore, if we look at Pareto, Pareto, for example, was very proud of his Theory of Choice in which he said that every psychological analysis is eliminated. And he boasts that he’s not interested in the reason why man is indifferent between one bundle of goods and another bundle of goods. “I notice the pure and naked fact.” And so words like willpower and imagination and feelings began to vanish from economics.
Instead, economists came up with very elegant and certainly very useful theories of well-being as preference satisfaction, where you’re happy to the extent that you can satisfy your consumer preferences, and you can approximate the extent that you can satisfy you consumer preferences by how much income you have. And therefore the recommendations of economists tended towards– both for individuals and for nations– well, do your best to get richer. Now, 20th century economists were not stupid. And of course, they’re not oblivious to the potential unintended consequences of that. But they have this special powerful Latin to protect their assumptions from that.
So we just say “ceteris paribus,” and that means all other things being equal– and definitely it’s true– but all other things being equal, aiming to get richer is a pretty solid idea. But around the 1970s, some economists started asking awkward questions on this topic. And in particular, here is Richard Easterlin of the University of Pennsylvania and now the University of Southern California, who’s commonly regarded as sort of the grandfather of modern happiness economics. And Easterlin published a paper in 1974 with the title “Does Economic Growth Improve the Human Lot?” which was probably the economists’ equivalent of is the pope Catholic and do bears go to the toilet in the woods? But Easterlin was sort of looking to confront this economic growth data with some kind of external measure of happiness. And in order to do that, he used self-reported well-being survey data.
And following Easterlin, lots of other economists have taken up this line of thinking. And one of the key figures, certainly, in this country is Professor Richard Layard, who is a labor economist at LSC and a member of the House of Lords. Layard wrote an influential book in 2005 based on some earlier lectures, and along with the former cabinet secretary here, Gus O’Donnell, has been kind of instrumental in getting the UK government to take well-being measures seriously, both within the civil service and in the country at large. I said Easterlin used self-reported well-being survey data. What is that data? Well, it answers to questions like this.
And actually, these are the four questions that the Office for National Statistics now asks regularly across the country of a representative sample. How satisfied are you with your life nowadays? And this relates to what we often call the evaluative account of well-being, thinking about your life as a whole. It asks what extent do you feel the things you do in your life are worthwhile? And that gets at what we often call the eudemonic account of well-being, sense of meaning, sense of purpose, sense that what you do in your life is worthwhile. And then the last two questions, how happy did you feel yesterday and how anxious did you feel yesterday, get at what we call the hedonic account, which goes back to Aristotle, as in fact the eudemonic account does too. That’s sometimes sort of rather deprecatingly referred to as smiley face feeling.
It’s kind of mood, emotion, in the moment, the most immediate kind of measure of happiness. Now, all three of these accounts of well-being are very highly correlated. You know, you’d expect that people who have enjoyable moments feel satisfied with their lives, for example, although there are some interesting divergences. Having kids might cause you to have an interesting divergence in some of these. You might get very high highs but also some fairly low lows.
You might feel that your life has more purpose. It might affect things differently. By and large they move together, but there are some things that can move them somewhat separately. Economists have mostly just looked at the evaluative account, because it’s really easy to get ahold of. It’s one question, and you can ask it today, and you can ask it next week and next month and you’ll get a similar answer.
So you can sort of shove it into any survey that you’re asking people. And it’s also the one that’s most closely related to people’s income, which makes economists feel comfortable with it. My own project, Mappiness, is actually primarily a hedonic measure of happiness. And not just how happy you felt yesterday, but how happy do you feel right now. And I will talk a whole lot more about that soon.
A key question that comes up in economics and elsewhere is very simple. Is this valid? Is this meaningful? Can people really reduce the quality of their life to 0 to 10 scale? And certainly within economics there remains an amount of controversy about this. There are certainly economists who are quite wedded to the preference satisfaction view, and to their indifference curves, and to not getting involved in the messy business of how people feel. And in reducing the quality of your life experience to a 0 to 10 scale, we can certainly be accused of reductionism. It is a kind of reductionist enterprise.
And I’m not for a minute saying that’s the only way that we should seek to understand people’s life experiences. But my view is certainly that the subjective well-being ratings are valid enough and meaningful enough to tell us interesting and valuable things. And for one thing, there’s a lot of convergeability. Lots of other things point very much in the same way. For example, subjective well-being responses are very strongly correlated with objective indicators, including neural activity from fMRI scans.
Biochemical markers, such as levels of cortisol in the blood, heart rate, blood pressure, smiling, other people’s ratings of a person’s well-being. They’re a strong predictor of people’s future behavior. And subjective well-being ratings are also very highly correlated with income. Certainly they’re very highly correlated with income if you look at a single country at one time. The income relationship is sort of less clear cut if you look between countries.
And we could talk about this all day. These charts are sort of interesting. So these are actually the same charts. One of them is on a log scale. One of them is on a linear scale. And the choice of which one of these you plot is actually potentially quite political. If you plot this one, you kind of make very clear that Robin Hood was doing a good thing, because up here, very little happiness distinction between people earning $100,000 versus $150,000. Down here a very large happiness distinction between people with a much smaller differential. So if you ever look at the Institute for Economic Affairs report, for example, they’ll always plot this one. But there is a strong relationship within a country.
Now, you do need to be careful when you’re comparing between individuals. And certainly you need to be careful when you’re comparing values between countries. But you can get a lot of mileage even out of just the changes within each individual’s own responses. And that’s actually primarily what my research relies on. So I would certainly argue that this is a valid and a useful way to look at the world.
Some of you may be saying, back up a bit, didn’t I see an fMRI scan a few slides ago? Wouldn’t that be the gold standard scientific way to measure happiness? You know, it’s got big machines, costs lots of money, spits out hard numbers. Give me a break from this hand-wavy social sciencey question asking stuff. But actually, no. Because happiness is fundamentally a completely subjective thing. It’s part of an individual’s experience.
The only reason that we know what the fMRI scans mean and we can interpret them in terms of happiness is because at some point in the past, we put people in fMRI machines and we asked them how happy they were while we were scanning their brains. So you can make quite a strong case that actually, the gold standard for knowing how happy people are is just to ask them. You might also remember that I promised that subjective well-being data could tell us valuable and interesting things. So what kinds of things would be? Well, there’s at least three ways that economists might want to use subjective well-being data. One of them is very simply to test and refine other economic theory.
There’s this kind of tautological sense in economics that what do people want is what they do. And what they do, well, it’s what they want. And you can’t really get out of that very easily in this kind of revealed preference, almost kind of behaviorist, way of figuring out what people want in their lives. But if you can measure their happiness, you can tell if people do seem to make the choices that make them happier, or whether they systematically mispredict what’s going to make them happiest. And there’s lots of good evidence that they do systematically mispredict, and they don’t make the choices that would have appeared to make them happy in retrospect.
You can discriminate between models that have different underlying utilities but look the same in terms of people’s behavior. Becker has an idea of rational addiction, where given a certain set of preferences, people can rationally become addicted to drugs. It’s very difficult to distinguish that when you look at people’s behavior from people just lacking self-control and not being in control of the choices they make. But if you can observe how they feel, then you might be able to discriminate which of those is in fact the explanation for the behavior that you see. And speaking today, having just been to your canteen, situations where people are happier having less choice, which is sort of paradoxical and impossible in standard economic theory, because more choice always kind of less choice is nested within it.
And more choice must be better. But it’s certainly the case that sometimes there are situations where we are happier having less choice, just one or two things to choose for lunch. Or I was in a situation where I got taken somewhere flying business class at one point. And it was amazing. And then the next time they offered to pay me the money, and then I had to make a choice whether to fly business class or not.
And that cost me an awful lot of anguish, because the price is ridiculous. And on the other hand, I kind of enjoyed it. So there are certainly situations where we’re happy having no choice, which don’t make sense in normal economics. But subjective well-being data might help you understand them. And there are probably more things than this you can do.
But the third of the things I was going to talk about are that we can value things and we can sense the importance of things that aren’t traded in markets. And this might be things that you really just couldn’t trade in markets at all, like marriage. And there is a paper that finds that marriage is extraordinarily valuable if you try and turn it into an income equivalent in terms of people’s subjective well-being. And also, for example, the environment, which in certain areas can be kind of traded in markets. And you can at least ask people to imagine that it’s traded in markets.
But nevertheless, that isn’t always very strong information on how people do value the environment and how it makes them feel. And that brings me to my research. And this is the logo for Mappiness, which was an iPhone app that launched in 2010 as part of my PhD at LSC. And this is, in a sense, quite a good day to take stock of this project, since actually the release of iOS 11 yesterday marks the moment when Mappiness just stops working more or less everywhere. The code was written for iOS 3. This was pre automatic reference counting, and pre almost everything else.
So the background. What Mappiness is for is looking at the impact– was originally, certainly– looking at the impact of natural environments on people’s happiness. Here’s green space. Same slide, I think, that was behind all the logos at the beginning. And we can all agree that green space, hopefully, is nice. It makes us feel good. And I once went to a geography conference where no one said anything else for three days. That’s fine as far as it goes. But it doesn’t go all that far, because what is actually kind of useful to know is just how lovely is green space, right? Policy makers want to spend money on all kinds of different things.
They can’t spend all their money on green space. And so it’s helpful to them to have some quantitative sense of the loveliness of green space, also blue space, other kinds of natural environment. So how much happier are people in natural environments? Economists looking at this have generally done big observational studies. And they’ve looked at countrywide averages of happiness and environment. Or they’ve looked at people’s general happiness on the individual level in relation to where they live and their immediate surroundings.
There are some problems with those approaches. So what I kind of set out in the PhD thinking was, wouldn’t it be nice if we could look at the experiences of individuals at particular moments in time? So not just how happy do you feel in general and where do you live, but how happy do you feel right now, and where are you right now? And that will give you some really direct data on the link between happiness and environment. And it would need to be that hedonic account of happiness, that smiley face feeling, because that’s the thing that varies sort of minute by minute and hour by hour. If you were to ask people how satisfied they were with their life as a whole in different environments, you wouldn’t probably get very much if you tried to do that every day. Now, looking at people’s momentary experiences has been done for many years in psychology and also in medicine.
In psychology, these are called Experience Sampling Method studies. In medicine, they’re called Ecological Momentary Assessment studies. And Alex tells me, actually, that some of you are running something very much like a Ecological Momentary Assessment study at the moment, using watches. But in the classic ESM survey, certainly when these things were come up with decades ago, you would give out notebooks or PDAs. You beeped people at random moments, and you asked people about their experience and the context of their experience.
And the benefit of this, you get no recall bias. You get an accurate and detailed record of experience that isn’t affected by the things that, for example, Kahneman talks about, about the peak/end rule and all of the kinds of biases. There’s an interesting philosophical question, actually, about whether you are more interested in what you experience or what you remember. But certainly you can make an argument that it’s nice to have no recall bias in this data. And we get panel data.
We get longitudinal data. And that means we can do fixed effects models, where we only use the variation within an individual. And that’s a big benefit. If you want to think about that benefit, here’s a green space in London, Buckingham Palace Gardens. And if we went and found that responses from people in Buckingham Palace Gardens were generally happier than the average, you might want to argue that the kind of people you find in Buckingham Palace Gardens are not quite the same as the average, and more likely to have just been bought a yacht by the taxpayer, for example.
There are all kinds of reasons why those people might be happier. But with the fixed effects model, when we’re looking only within individual variation, we can tell that the same people in Buckingham Palace Gardens are happier than when those people are on the other side of the wall facing lots of buses driving past them on their bicycle, for example. And so this is a kind of powerful way of looking at this, because we get to use that within individual variation and take out effects that say that certain people are found in nice environments and those are different people than go elsewhere. And then around this time with the PhD, the iPhone launched. And then apps became available for the iPhone.
And this seemed a great tool to us. So ESM studies began in the ’70s, and they were kind of difficult to run, because you had to give out these diaries. You had to give out these pagers. And they found that in some studies, almost half the people filled in all the diaries at the beginning to get it out of the way with, or at the end, because they realized they haven’t done it yet. And it’s a real logistical challenge, whereas with the iPhone and now with other similar devices which I’m told exist, you’ve got GPS.
You’ve got a microphone for sound levels. You can send the data back using a mobile data connection. You can interrupt people with a noise at any moment. You can even have them take a picture of the environment. And of course almost everyone now has one of these things. And there was a problem, which is that certain kinds of people have iPhones. And I suppose certain kinds of people have Android phones too. Certainly in 2010, iPhone owners are on average sort of young and rich and educated. And they have to self-select into the study. And then they have to be able and willing to respond to a beep.
So certainly, we don’t have anything approaching a representative sample. And you can grill me on that later if you want to. But if I do want to take part, then 2010, 2011, 2012, which is when most of our respondents are for, it’s a free download in the app store. I spend two or three minutes signing up. And then I get dinged.
And then I spend sort of 30 seconds or so responding to a series of questions. I say how happy I am. And that’s mostly the data that we use. That’s mostly the dependent variable in the regressions that I’ll tell you about in a minute. And then I report a few control variables which sometimes turn out to be quite interesting in their own right.
But I say who I’m with, what I’m doing I think at this point that I’m imagining that I’m in a pub. And this will relate to one of the things I talk about later. And then all the data gets sent back. I can look at my data in the app, get some very simple information. I think actually just being caused to reflect on your happiness twice a day is more powerful, probably, than a few simple charts. There’s also a sort of adjacent API where I can build things off this. And there’s even a kind of an online section where I can look at my data on the Google map. I can view charts. I can look at all my pictures, and so on.
And there’s me working on Mappiness website, I think. The project got really good media coverage. I was sort of thinking, this is a PhD. I might get 100 people doing it for two weeks. That would be fine for a PhD thesis. But actually, we launched in August, which was a great idea, because nothing happens in August. And so we got really nice coverage, interview on CNN, front page of “Le Figaro,” not quite so useful because we’re not targeting France. But anyway, Tuesday is worse than Monday, and that’s a really big story. Featured in the app store, so that now we have more than 4 million responses from about 66,000 people. And if I give you some sort of very simple descriptive, this is the first year of data and a half hour moving average of how happy people report being.
It’s a little bit washed out at the top, but the left I think starts at 8:00 AM, and the right ends at 10:00 AM. And then there’s a sort of horizontal line per day through the year from the beginning of September 2010 to the end of August 2011. And a few patterns might jump out at you. Firstly, it’s stripey. And the stripes are weekends. So you can see lighter colors are happier. Weekends are enormously happier than weekdays. There’s a kind of really bright spot around here, which is Christmas. And there’s a really bright spot around here which is sort of the May bank holidays. We get a little bit lighter as we go from the left to the right, which tells you that, on average, people get happier throughout the day.
And you can also see that the second half is a little bit noisier than the first half, so you can tell that we had slightly more respondents in this period than we did in this period. This actually now goes over the whole 6 and 1/2 years up to the beginning of this year. It’s probably too small to see, which is sort of vaguely intentional, because I still haven’t published this. But you can see these peaks are all Christmases. Christmases are always good.
There’s a general downward tendency. I don’t know whether we should make anything of that or not. It might just be that different kinds of people were signed up later than were signed up earlier. So maybe some face validity for saying that people are a little bit less happy now than they were in 2010? But I couldn’t guarantee you that from the data. A few other events that we can spot. Murray winning Wimbledon here on a hot day. Hot days of above about 25 degrees, you get a really big spike in happiness. Other weather variables make a difference, but not such a big difference. But really hot weather is very good Olympic opening ceremony, royal wedding, annoying for Republicans.
And you probably can’t read this, but it relates to one of the questions. This is the worst, the least happy day that we recorded over the whole of the 6 and 1/2 years. And this is the news that Trump got elected as the president of America. This is, I think, the Brexit result. I think that’s the fourth least happy day.
.Now, you might say this just shows you that Mappiness users are not representative of the population. On the other hand, referendum voters are not representative of the population either. And Mappiness voters have great taste. So I seem to have made a more political point than I intended. Planning to use this to create a happiness forecast at some point.
And I’ve been planning this for a long time. But there’s this idea of Blue Monday every year. It’s the second Monday in January. They kind of wheel out the guy who came up with Blue Monday. And sometimes they wheel me out to say that Blue Monday is rubbish. But we’ve got no evidence that the second Monday in January is anything particularly bad. If anything, it looks like a Tuesday in November might be the worst day of the year on average. But I hope to have news on that in the near future. Now, in Mappiness I was interested in green spaces. And so as an incredibly naive first approach, this is measuring the number of green pixels in each of the pictures that people take.
These are the colors that I’m counting as green. And this is some indication of how many pixels there are in each one. It was done very simply in processing. Interestingly, the amount of green is very strongly correlated with happiness before we control for anything else. So no green pixels to all green pixels gives you about six points out of 100 jump in happiness.
I’m not sure we can read very much into that, because the effect goes away when we control for other things. It might well be, for example, that people really don’t like work and that workplaces generally don’t have a lot of green in them. And it could be as simple as that I’ve also been thinking about deep learning for image classification. There’s about 300,000 photos, so that could be fun.
I’m sure there’s lots of other kinds of deep learning that can be done with this data. But for the PhD, and the more serious analysis in the PhD, I joined up habitat data with each response location. And this is an extract from the Land Cover Map 2000, which is a high resolution remote sense satellite data set. This is London, so you can see a lot of black, which is continuous urban. But you can also see the river.
You can also see Hyde Park as a mixture of trees and grassland and so on. So this is not about country versus city. It’s very much about land cover wherever it occurs. And then all the information that we collected goes into one big regression model. So it predicts happiness from naught to 100 on the left, with on the right the habitat that you’re in, the weather conditions, the daylight, who you’re with, what you’re doing, what kind of place you’re in, time of day, day of the week, how many responses you’ve given before, because we find that people get happier over time, and we use individual fixed effects to take out that average effect of being each person.
And these all go in together. And that means that each effect is controlled for every other effect. That’s important, because when you’re out in a green space, it might also be sunny, and you might be playing Frisbee with your mates, and it might be the weekend, and so on. So this all goes together to this big regression model. To get a feel for the size of the effects, you can look at the happiest and the least happy activities.
The top activity is one that I added after many emailed requests. There may be people looking over people’s shoulder at this point, so we take that one with a pinch of salt. But you can see that generally it’s sort of physical activities and cultural activities and social activities that seem to occupy the top spots. And most of these give a give a bump of sort of four to six points out of 100 in people’s happiness. We’ve got plenty of data, and so these are the 95% confidence intervals.
So these are very significant effects. And then the focus of the research, which was land cover. And we find this land cover type, marine and coastal margins. People were enormously happier in marine and coastal margins than anywhere else. This is all relative. Zero here is being in an urban environment. So people are on average about six points happier in what’s effectively an estuary. This is a kind of an estuary environment, marine and coastal margins. It’s not any coastline. But in any other natural environment, they’re also significantly happier than they would be in an urban environment after we’ve controlled for all the other things that we can think of.
And this stands up to a number of different robustness checks and some completely different ways of defining a natural environment. So it works very similarly if we use areas of outstanding natural beauty, national nature reserves, and one other the classification that I forget now. But different kind of administrative classifications of countryside give us sort of similar sized results. And people are happier in those than they are outside. And so this was reported in a paper in the journal “Global Environmental Change.” A strong line of evidence on links between environment and happiness. Slightly ambitious title. This is only the UK. It’s not representative. But anyway, we’ve got some nice quantitative evidence on this point.
And that’s nice, because we all probably, as I said at the beginning, realize that people like green spaces. But this generally sounds a little bit soft and intangible and perhaps doesn’t stop the government proposing to privatize our forests or allowing them to bulldoze sites of scientific interest for golf courses and so on. And so the more hard quantitative data we’ve got on this, perhaps the better the chance we might have in saving some of the environmental treasures that we’ve got. So we’ve covered being by an estuary. That’s really good.
And we’ve covered Donald Trump winning the election, which is very bad. We’ve got a number of things left in this list that I promised you. Any of you that are werewolves will be interested to know that I don’t find any effect of the full moon or any other phase of the moon, although I do have to control quite carefully for other time variables in order to not get interesting false positives on that point. And that leaves us with drinking, riots, and football. Unsurprisingly, we do find that alcohol, drinking alcohol, makes people happier in the moment than when you’re not drinking alcohol.
And that’s after controlling for lots of the things that go along with drinking alcohol, such as probably not working, being with friends, being in a pub, and so on. And we also see it making you tired the next day, which is not rocket science. These were back of the envelope calculations for the cabinet office who were looking for evidence of the benefits of local pubs as sort of community hubs. These, again, are fixed effects regression results controlling for a whole range of other things. And we don’t know what people are drinking.
We don’t know how much they’re drinking. But we do find a big positive effect of drinking, six points out of 100 in happiness. And that’s very significant. And then we have some other interaction. Drinking in the pub is better, drinking on your own is significantly worse, and so on.
Now, that’s the basis of another paper. And, of course, that’s not the whole of the well-being story with alcohol, right? We saw negative coefficients on drinking alone, drinking on a weekday morning. There are, of course, drinkers and non drinkers who have their lives wrecked by alcohol. But the positive effects are part of the story. And actually, some recent research work and some recent policy work often can entirely neglect to mention the positive reasons why people drink as well.
So it is useful to be able to quantify those. And that’s the point that my coauthor Ben Bamburg makes in this paper. I have very little interest in football, but some of my colleagues at Sussex are football fanatics. And it seems to be a source of strong emotion for some people, judging by these quotes that I’ve picked out. And so we figured out that actually with the Mappiness data, we can look at this.
And we have responses like this one, which occurs here, kind of very close to the football stands, middle of the afternoon on a Saturday. And the individual reports that they’re at a match or sporting event. So we found a data set with the locations of football stadiums. We found another data set that has match dates, results, and the bookmakers’ odds. And we did a little bit of web scraping to find the exact kickoff times of those matches.
And then we could join all that together to figure out who seem to be keen football supporters, which teams they follow Ideally you spot them at two different matches with one team in common and then you’re reasonably certain. And we get 100 or 200 people that way. And we get a few hundred more if we just assume that people are probably supporting the home team. And then we can follow those people’s happiness level throughout the season.
We’re not limited just to the matches that we spot on the map. And this is actually a paper that I’m working on right now. And what we find is interesting and very much in line with sort of the behavioral economics on loss aversion. So we find that a win is sort of niceish for an hour or two. A loss is really bad. It’s much more negative than a win is positive. And not only is it much worse, it also lasts much longer. So you can still see the next morning that people are unhappy if their team lost, even more so if they didn’t see it coming. So if you also have an interaction in there for the bookies’ odds being that they were going to win, and then they lost, that makes it even more painful. And actually, this does have some quite interesting sort of real world outcomes as well.
There was a paper in the “Quarterly Journal of Economics” about American football and looking at domestic violence that came out with very strong findings. And we may also be able to look at people’s behavior. We haven’t done this yet, but kind of quite interested to find out whether it has any effect on the probability of showing up at work the next day. So that leaves us with riots, I think, of the things that I promised to talk to you about. And this is work that was done by my PhD student [Panka] who wanted to be here, but unfortunately isn’t quite back from the States on a Fulbright scholarship.
This follows on from some work where we looked at happiness impacts sort of rippling over time and space from terrorist incidents in Northern Ireland. So [Panka] took the Mappiness data and was looking for a sort of signature from the riots that happened in August 2011 in English cities. And if you weren’t here or you don’t remember, there were sort of several days of looting and setting fire to things and clashing with the police that started, I think, in Hackney, and spread to other parts of London and then the country as a whole. Now, that seemed reasonably promising to me, because I was living close to Elephant and Castle at the time, and I remember that we were planning to go out to a restaurant and then we thought maybe the Elephant and Castle shopping center isn’t the best place to be at this point. So I thought you’d have seen it in my happiness data.
[Panka] hasn’t actually managed to find this in the Mappiness data, which I don’t think is her fault. I think we just don’t see it. All that we do see is that we can see that people are watching telly a little bit more over this period, which is potentially people keeping track with what’s going on on the news. We’re not quite sure. But as far as we can tell, these riots don’t have an impact on your average Mappiness user.
So that’s all the things that I promised you, but how are you doing today? Well, the bad news, obviously, is that you’re at work. And of all the things that you could be doing, work ranks second only to sickness as a cause of misery. These are all the activities that people can report doing in Mappiness. There’s about 40 of them. And here is working or studying.
And we can restrict this just to the people who say that they’re in full time work, and we find that has quite a big negative coefficient, second only to being ill in bed, which is rather bigger. That’s 20 points out of 100, but work is still minus 5. And in fact, it’s not just being in the workplace. The working day and working are independently negative. I can’t remember how colleagues turn out.
Maybe I shouldn’t say anyway. This is work with Alex Bryson at UCL. And we also look at what you’re doing while you’re working. And we find, in fact, that only 03% of people follow Snow White’s advice, which is a pity, because the happiness effects of singing and performing would be just about enough to wipe out the negative effects of working, although the effects on your coworkers might also need to be factored in there, I suppose.
So there were two other things that I want to mention today in terms of what we’ve been doing and what we’re going to do next. And I mention these two things particularly just in case anyone here is sort of interested and potentially could help out. Now, this is a bit washed out. And so my PhD student, [Panka] is now sort of turning her attention to air pollution. And we’ve got really good reason to think that air pollution will affect people’s subjective well-being, including routes via their physical health.
We’ve also got really good reason to think that it will impair their physical and cognitive abilities. There’s a really amazing paper last year from LSE by Sefi Soth, where he gives quite convincing evidence, having kind of put air pollution monitors in the exam rooms at LSE, that high levels of particulates in the exam room increase students’ probability of failing their exam by about 5%. And he has the same students sitting different exams in different rooms at different levels of pollution so that the identification is quite convincing in that one. Now, I raise this possibility of a link between air pollution and performance because one thing that occurs to us is that Google has really rich data on some kinds of cognitive and perhaps also physical performance, for example, maybe mistyped searches, mistyped words on Android keyboard, and so on. And these data presumably could be located in space and time.
And that means that potentially they could be related to air pollution data to look for potentially causal relationships. And in cities like London and lots of American cities, different places around the world, there’s quite good data now on air pollution through space and over time. So if you can think of any data that would help us understand air pollution effects on performance, then do let me know at the end of today or on email or via Alex. It doesn’t necessarily have to be big things, you know, thinking about a simple typo. A simple typo has been enough in various situations in the past to administer a near fatal drug dose.
To lop several nines off the availability of S-3, we’ll have the world believing that there’s an order of magnitude more iron in spinach than in fact is in spinach, which is the origin of the Popeye cartoons. So even small things, I think, could be really interesting there. And the other thing I was going to talk about, which as I say is sort of apposite on the day after Mappiness kind of goes dark, is Mappiness 2. And there are sort of three things that distinguish, we hope, this next iteration of Mappiness, which I’m already working on, from the original. Firstly, it will be on iPhone and Android, and we hope it will add in some fun stuff like heart rate variability measurement with the camera and flash which, I discover, Alex and some of his team know an awful lot about.
Integration, probably, with HealthKit and perhaps equivalents on Android, which I’d be happy to talk about. It will try to not only quantify your happiness– and as I said, we do find that the process of reflecting through doing Mappiness for a while does seem to make people happier– but also to improve your happiness. Now the idea in the original Mappiness was just that we wanted data, so we kind of didn’t want to change people’s happiness. Obviously it’s a good thing that people’s happiness went up rather than down, or we’d have a lot of trouble getting ethical approval for any future study. But it’s a little bit cold in that you can tell Mappiness that you’re really miserable day after day after day, and Mappiness the original does nothing about it.
My co-founder at Psychological Technologies, Nick Begley, was the first employee at Headspace, and so he knows plenty about mindfulness and other sorts of scientifically validated approaches to making people feel better. So the idea is that in the next Mappiness, if you’re not feeling good, we can begin to offer you suggestions for the kinds of things that might make you feel better. And for example, we have mental and physical well-being interventions that are integrated in Mappiness 2, like this sort of interactive mindfulness practice. And we can suggest them at appropriate moments. And even better, hopefully over time, we can collect data on what the most appropriate moments are for which kinds of interventions and how those are going to make people feel better.
And the third thing that’s new in Mappiness is more of a focus on work. So as I said, the first Mappiness found that people aren’t happy at work. And that wasn’t what we set out to do. And we therefore didn’t collect lots of the data that you’d probably most want to try and dig in to that in more detail. You know, we don’t know what sector or even what company people are working in, what they do at work, anything about what they think about their employer, and so on.
And this is obviously information that employers are interested in and that employers could use to make the workplace better and make people happier at work, as well as perhaps more productive. So we already have employees in one big household name firm using what’s effectively a kind of an alpha version of Mappiness 2. But we’re also looking for people, and indeed workplaces, who would like to help us conduct some kind of randomized controlled trial on the impacts of this sort of technology, measuring and then trying to improve people’s well-being through the apps in their pockets. So that’s it, but hopefully there’s plenty of time for any questions, if you’ve got any questions. And thank you for listening.
SPEAKER 1: We do have time for questions. Please raise your hands. Yes.
AUDIENCE: Hi Quick question. Have you looked at the cultural or ethnic component? Does this apply to everybody of all ages, and things like that?
GEORGE MACKERRON: That’s a very good question. We don’t ask questions about ethnic origin, and we’re primarily looking in the UK, so unfortunately I don’t have much data on that. We do ask some demographic questions, age, gender, marital status, employment status, and so on. Now, generally, the analysis that I’ve shown you here have been these fixed effect things. And so all these individual level things drop out, and we’re just looking at the average across everyone.
We have looked a little bit, I think, at some gender differences and some age differences. And in a lot of cases, there isn’t any single thing significant, and in some other cases, there’s not very much I’m a little bit hazy now on the answers. Now, there’s definitely scope to do more there. You could do more with the data that we’ve got.
You would ideally, actually, ask some more questions as well I mean, the focus on Mappiness, really, was let’s look at natural environment, and let’s make it so quick and easy that possibly some people will want to do it. So in retrospect, there’s quite a lot of things that would be nice to look at I mean, do you have any hypotheses about what you might find in any of that?
AUDIENCE: No, I mean, there [INAUDIBLE] some cultural components to how somebody perceives drinking in some cultures versus how somebody perceives drinking in a cultural context. And that can affect the—
GEORGE MACKERRON: No, that’s right. So I guess results that I show you today are kind of valid for the subset of people in the UK that we’ve got. And you’re right, actually. That’s certainly a good example where you would have some different effects, I’m sure.
SPEAKER ONE: More questions? Yep.
AUDIENCE: Is there any more you can say about people seemingly being very happy at Christmas? Because my experience is that everyone moans about how terrible Christmas is.
GEORGE MACKERRON: So, I suppose a lot of it turns out, I think, to come back to how likely people are to be working. And I suspect that happiness at Christmas is partly driven by the fact that we’ve got the least number of people working throughout the year. We see bank holidays as being very positive as well. Beyond that, it’s just anecdotes, isn’t it? I mean, it would be good to do qualitative research in some of this stuff. And I guess that’s not my area.
There’s perhaps a slight worry about whether people don’t report if they’re not happy at Christmas because they feel like they should be, or if they misrepresent their happiness at Christmas, again because they feel that that’s the time that they’re supposed to be happy. They don’t want to admit that they’re not I mean, if you were going to find that in any kind of well-being research, hopefully you wouldn’t find it in this, which is a very kind of private thing with you and your phone. It’s not someone ringing you up and asking questions. But yeah, it’s a good question.
People have sometimes sort of expressed surprise at that, and sort of said, hmm, people spending a lot of time with their family isn’t always a recipe for high happiness. So, yeah, it’s a good question.
AUDIENCE: I mean, I think the obvious question is are people otherwise happier when they’re spending time with their family?
GEORGE MACKERRON: And they are. Yeah. But I mean, the things that we looked at where happiness was amazing, that was not controlled. That was just sort of raw averages. So you’re right that if we sort of start taking out the variation that we can predict, then you will see those peaks drop almost certainly, right? Because they’re with their family. They’re drinking Probably some of them are drinking. And they’re doing other things that they enjoy. And they’re not working. Yeah So some of it’s—
AUDIENCE: –they’re not lying.
GEORGE MACKERRON: Right OK. So that’s certainly in line with– yeah Yeah. And I mean, it’s an average effect, so you would need, I guess, as long as 2/3 of the people are happier, 1/3 of the people could be less happy, and the average coefficient here comes out the way that we’ve seen. So there’s also certainly scope here for looking more at the distribution, and maybe doing quantile regression and seeing what drives the low responses versus the high responses, say. Always in these talks there’s a million things that I have to apologize for not having done. There’s a lot still to do. Yeah.
AUDIENCE: So with the fixed effects model, you’re sort of looking at differences within individual people. Like they’re happier when they’re in a green area. But I mean, is there some concern that people might adapt to these interventions? Like maybe the government puts green spaces everywhere and there’s still a big difference, but instead of being really happy in green spaces and just kind of meh in city urban environments, now I’m used to green spaces and urban environments are just really terrible? Or bitter, because of the difference.
GEORGE MACKERRON: Yeah. And I guess related to that, there’s a question of sort of contrasts. And so most of our respondents live in cities. And so maybe it’s partly that they’re getting out of the city and they’re going on holiday. And that’s actually the number one question that I wish we’d asked, is are you on holiday? And we can kind of try and get to that with various heuristics, but we don’t quite have it.
So that’s an obvious thing to try and get at, also looking at people who perhaps live in the countryside and then visit the city. That’s sort of one of the top things on my list to stop apologizing for not having done and do. And I promised a book chapter later this year, actually, where my plan is to do that amongst some other things. In general in the happiness research, sort of habituation or adaptation is certainly an issue. And I mean, one of the things that drove me to do this was I moved down to London to do the PhD. I was previously in Cambridge at that point. And Cambridge was green and quite clean. And London was incredibly kind of dirty and polluted. And that kind of got me down. By the end of five, six years in London, I was getting– at least, I was aware of it getting me down less.
So no, that is a question. And that is certainly a question for policymakers. You know, there’s all kinds of other things people adapt to as well. You know, they adapt to sort of chronic health problems that the government thinks it’s worth spending quite a lot of money to avoid. Those then don’t seem to affect people’s happiness in the very long run as much as you’d expect. One of the reasons for that might be that happiness isn’t the only thing that people are interested in. And in fact, the football research also talks to that idea that happiness might not be the only thing that people are interested in. Because unless everyone supports teams that always win, which I’m pretty sure isn’t true, then losses are so much more painful than wins are pleasant that you’d think that they weren’t maximizing the sort of the integral of their happiness by going to football matches and following them. So that’s a bit of a rambling answer, but it kind of—yeah. Your point is still valid and interesting.
What you’d, I guess, like to do is have some kind of, again, some kind of trial, some kind of random assignment of more or less green space. See what happens if a park ends up being closed for a year to the people who live nearby, say. And then you might look at the profile over time. If there is some adaptation, how much is there? I suspect that it wouldn’t be complete. There are lots of other things that– the reasons that we think people like green spaces, you would imagine perhaps wouldn’t go away.
So you know, there’s an opportunity for sort of physical exercise that improves health. There’s an opportunity for socializing with people. And there’s also this idea that we evolved in the savanna, or on the edge of forests, or wherever it is that we evolved over tens of thousands of years. It might well be something that just kind of causes us to find these patterns in our visual system and kind of attractive. You this biophilia hypothesis of Eo Wilson. So I would be surprised if all of those were totally adapted away. But it would certainly be interesting to find out how far they were.
AUDIENCE: There’s a new park in East London, right, during the time period that you studied?
GEORGE MACKERRON: Yeah, the Olympic Park. Yeah, we should look at that. We have briefly looked at the Olympics, mainly with the ax to grind that. I wonder if it would have been an awful lot cheaper just to have a couple of extra bank holidays. Might not turn out to be true. I haven’t really checked in enough detail. I suspect that we don’t have enough responses from that time. And there aren’t enough people living around it. Because they’re also kind of building– well, I guess the houses have just been built, haven’t they? People have presumably all just kind of moved there as well, because it was a sort of big redevelopment. It’s an interesting thought. It’s the right kind of thing, I think, for us to think about looking it. Yeah.
AUDIENCE: Hi, sorry. You had sports and doing sports as one of the top things that drives happiness. But also drinking alcohol, as well, was on that list. Just wondering about the trade off between like virtuous behaviors and then maybe not so virtuous behaviors and how they impact happiness.
GEORGE MACKERRON: Yeah, I’m not sure I have anything very intelligent to say about that, apart from that, yes, it does sound interesting. And I suppose if we’ve got people that we observe over a long period of time, then we could potentially, I suppose, look at that kind of cumulative effect of what you’ve been doing. I mean, one of the things that actually I don’t do at all here is you don’t have any dynamic models. So we kind of account for the responses being from the same people in the era.
But we don’t use the fact that we could kind of, this next response follows the previous response. And so, in general, we could kind of both do that on a kind of shorter level and I guess also on a longer level, and see people who often report that they’re doing sport versus people who often report that they’re drinking, for example. Some of these things, certainly the drinking, potentially has reverse causality, because it might be that people drink because they’re stressed. And it can be awkward, but with the time series data, we might have a chance of addressing some of that, because you can see what precedes what.
AUDIENCE: It seems like most of your data and results here are from local areas. Do you have any plans with the new release to target the US, Korea, other industrialized areas?
GEORGE MACKERRON: Right. So the other thing that I should have said about Mappiness 2 as well as being on more platforms and doing a few more things is that, yes, it should be worldwide. And it would, I guess, have been a good idea if Mappiness had been worldwide. And I have to kind of– it was quite late in the PhD, and timezones were really screwing with my head. So you can use it.
You can use the original Mappiness anywhere in the world. You just have to adjust the times. You’ll be beeped in London time. You know, it’s a silly thing, and probably would’ve taken me a day or two and I could have fixed that. And then we’d have had more people. But I guess the other question is I couldn’t– you know, what I was planning to do, the key thing was to join this with environmental data and do research. And I couldn’t promise for 100 countries in the world that I’d go and find environmental data and make that join up. So I couldn’t have ever said to people, you’ll definitely be part of the research. But yeah, definitely it will be good. And that would also, I guess, speak potentially to this question of different cultures and so on. And that would be interesting to do.
AUDIENCE: Here at Google, we have a whole people analytics team. Have you spoken to them?
GEORGE MACKERRON: No, but that sounds good. Yeah, put me in touch. I’ll talk to you afterwards. That would be great.
AUDIENCE: I’m sorry if you mentioned it I missed it. But did you ask gender question? Any correlations there?
GEORGE MACKERRON: I haven’t gone into the gender data in lots of detail. In terms of natural environment my recollection is that there’s one small thing that was slightly at odds, but largely the nature effects were kind of fairly consistent between the genders. Yeah, we do have that. As I said, most of these have been done with fixed effects, where you only look at the individual variation. So all the effects that pretty much don’t change for anyone aren’t seen. And in fact, we don’t give people an opportunity to tell us that any of those things have changed, either.
AUDIENCE: Did I see that the app can– it prompts you for a response, but you can also self decide to respond? Are there any differences between these prompted and self-reported?
GEORGE MACKERRON: So mostly, I mean, there’s an advantage in letting people do it at any time, which is that they feel like they’re in control, and they can also sort of show it to people and show them how it works. It’s a good question. It’s not something I’ve focused on. So generally, we’ve just taken the responses that were prompted, because the idea of this experience sampling method is you want a kind of probability sample of all the moments in someone’s life. And of course, they may have reasons. They may be showing it to someone, or they may be doing it because they’re particularly happy or particularly unhappy.
So I suspect that you’re right, they would be different I guess I’m not 100% sure what the difference would tell us, but it would be a thing that we could look at. Generally we can take a response, I think, that’s within an hour of having being signaled. And about half of signals get a response within an hour. You can reduce that or increase it, I think. And in ESM work, typically, it’s sometimes been just 20 minutes. But then in ESM work, typically people have made up half their responses. So it’s kind of swings and roundabouts. It doesn’t make a lot of difference to the effects that we were looking at here, whether you take a really short window or a really long window, which is encouraging. Of course, we are missing half of people’s responses.
And in fact, the original Mappiness requires a data connection to beep you. It was push notifications, because local notifications weren’t available. Which is a shame, because people up in the middle of nowhere up a mountain might be the happiest, but we can’t ding them. But yeah.
AUDIENCE: Have you seen anything interesting as people have aged through your study?
GEORGE MACKERRON: No. I guess we could try that. It would be difficult, though, I suppose, to disentangle– I don’t know if we’ve got a long enough time series to disentangle you aging versus the world going up in flames. And I mean, in general, we’d expect a big U shape. So in the subjective well-being literature generally, there’s this kind of big U, where being 40, 45 is the worst. And then it picks up again.
And some of it picking up again is that, again, people retire, and they’re not working anymore. But that’s what we expect to find. But I haven’t looked at that in Mappiness. As I say, I’m not sure if we’ve got the best data for that.
AUDIENCE: You mentioned some of these scenarios where there’s a lot of variability in terms of happiness for a single event, for example having children. Are there things that are much more consistent, and then so maybe the government could do or companies could do to improve? Are there things that everyone agreed on as being happy things?
GEORGE MACKERRON: I mean, one of the interesting things in relation to economists normally saying, just get richer, versus us thinking are there unintended consequences of getting richer? I mean, certainly one of the things that springs to mind is sort of people in city jobs who get paid vast salaries but don’t have any option at all except to work full time or double full time. I kind of think that working time regulations are potentially quite promising. Now, you know, if you could have people cut their working hours, but not have it be a signal that they weren’t committed, they might do well with four days instead of five days and 80% of their salary, say. The Economics Foundation have quite a lot of interesting work about that, and that potentially is also kind of good for the environment, because, you know, your city lawyer only has a sort of one weekend in every four off, and then jets far away and comes back. Probably isn’t made very happy by that, and in addition, they’re kind of using a lot of the world resources to do it.
Work life balance issues, potentially. Yeah, I don’t have a brilliant answer to that. I mean, redistribution, potentially I showed you this chart about how little difference it makes kind of up at the top end of income, how much more income you get compared to how much difference it makes at the bottom. And economists are normally very worried about that. They’re like these Pareto improvements, right, where everyone can agree that if you can make some people happier and no one less happy, then that’s great. But they generally kind of don’t like to talk about the ones where you make someone a tiny bit less happy to make someone else much happier. And you know, that’s a difficult political thing. But that’s some of what politics is about, I suppose. So there’s a thought. Actually, this is both my Twitter handle and GitHub handle. And probably not worth going to the Twitter, because it’s just political ranting.
AUDIENCE: So I have a question. Which of the findings of the study have surprised you the most?
GEORGE MACKERRON: And this is the question that journalists always ask. Yeah. Which of the findings have surprised you the most? I don’t know I mean, I guess I’m kind of used to these findings now. And the thing about most of them, you probably noticed a couple of times I was apologizing that they were kind of obvious. But the thing that it’s nice is to quantify them and compare them to each other and show that they’re real. So there hasn’t been much that has kind of come completely out of the blue. No, I’m not sure.
SPEAKER: We have time for one more question. Any last question? If not—
GEORGE MACKERRON: I mean, to be honest, the thing that surprised me most is that 66,000 people wanted to do this, on average, for six weeks at a time, which was wonderful. And I think they did that because they learned about themselves, and they felt like they were contributing to a sort of research enterprise. So that was good.
AUDIENCE: So this is my second question, so I feel like I’m cheating a bit. Have you considered talking to someone like Facebook, who with a single slider on the front page saying how happy are you, could get you billions and billions of data points very rapidly?
GEORGE MACKERRON: Yeah. And during the PhD I did think about kind of Facebook apps and that kind of thing I haven’t approached Facebook. I know, actually, that Facebook worked a little bit with the guy behind trackyourhappiness.org, which is a somewhat similar study based in the States. I think they have slightly fewer data points, and they don’t have location. But otherwise, they had quite a sort of similar approach to things. And I think he actually works part time there. It could be worth thinking of.
I suspect that an awful lot of researchers are kind of sidling up to Google and hoping for that kind of thing too. So I don’t know. But yeah. And I guess you’d get this kind of social network, potentially. You could sort of trace through some of the way that it works through networks, too, with that data, which obviously would be fascinating.
SPEAKER ONE: We are out of time. Thank you very much, George MacKerron.
Related Posts
- How to Make 2026 the Best Year: 6 Questions to Ask Yourself – Mel Robbins (Transcript)
- Dr. Karl Pillemer on The Mel Robbins Podcast (Transcript)
- Transcript: Denzel Washington On Legacy, Wife & Purpose – NXT Chapter Podcast
- How to Treat Men Better – Alison Armstrong on Modern Wisdom Podcast (Transcript)
- Why Young Men Are Struggling Right Now with Oprah & Scott Galloway (Transcript)