Lecture 22: Happiness and Mental Health

• Download English-US transcript (PDF)

[SQUEAKING]

[RUSTLING]

[CLICKING]

All right, I'm going to get started. This is lecture 22 on 14.13 on happiness and mental health. OK, so what are we going to talk about today? Happiness and subjective well-being and a little bit of mental health. So we're going to talk about rationality and revealed preferences and what can we learn from choices, and what does that tell us about happiness. We also talked about utility and how we can measure utility. In particular, how do we measure happiness, and is that a good or bad idea?

And then at the very end, we'll have, hopefully, a llama or a goat visit, which will at least improve my happiness, hopefully yours too. I've had some trouble communicating by email with the llama, goat, and their keepers. So I'm not exactly sure when they're going to show up, so they're scheduled-- the problem is, so they had only different certain time slots. The only time slot that was available was at 2:30, so I asked them to show up earlier, 2:20 or so.

I hope that's actually going to happen, but I haven't heard back, so I'm not exactly sure. So it might have to be at like 2:30. I hope you don't have to run. There was also some confusion about which Zoom room code the llama should go into, and hopefully, they find our room, even though they don't have and MIT ID. So anyway, that's that.

And then on Monday, we're going to talk about policy with people agents, which is very much the idea that, well, now if people are potentially making mistakes or have certain biases and we might think that we know about it, then perhaps we can do policies to improve their behaviors. And in some cases, that's straightforward. If you think people have biased beliefs or just have wrong information based on which they make information, then you might say, well, it should have just improved their information, and then they make better choices.

In other situations, it's much trickier, because A, people might have certain preferences that are not stable over time, and then it's very tricky to figure out what is the objective function that we should maximize. Second, it might be tricky because who knows what we know versus what people know? So who knows that the government or anybody else knows actually better what's good for other people?

So in some sense, if you want to intervene, you better be reasonably confident that what you're doing is actually well founded and you know better what's good for people than they know about themselves. And that gets you in tricky situations, because potentially, if you get it wrong, or maybe if you just misunderstand people's preferences, or their choices, or behaviors, then you might intervene and make things actually [AUDIO OUT] because you just misunderstood [AUDIO OUT].

OK. So what is rationality in classical economics? How do we [AUDIO OUT]? So broadly, we think that beliefs, preferences, and actions are rational if they are mutually consistent. What do we mean by that? Well, we mean by that that if we see certain behavior or certain actions, we can measure people's beliefs in certain ways. And then we essentially can find preferences that could rationalize these actions.

And so that's a very odd definition, and without that definition, we can essentially say it's possible to be a rational cocaine addict. It's possible to rationally commit suicide. It's possible to rationally marry somebody who you met six hours ago. And it's also possible to be rational violent offender. If there's preferences that people have, you could write down some preferences that rationalize all of those things.

And so [AUDIO OUT] in mainstream economics very much a main assumption. So the researchers' jobs is to identify preferences that are consistent with observed human behavior. So when you see people [AUDIO OUT] way, so then the question is like, what are preferences that rationalize that behavior?

Now what's the idea? We talked about this already before. So actors make choices, economists observe their choices, and then [AUDIO OUT] generate these choices if an actor were perfectly rational. [AUDIO OUT] tend to give these imputed preferences normative meaning. That's essentially, if I know your preferences, that [AUDIO OUT]. And that's what you should be doing, as in, if you prefer B, I should be helping you to do that. That would be a good thing for you.

Now just to give you an example of where this could lead you potentially astray, or where you might argue, is that really rational versus not, for example, take Jack, who prefers taking cocaine over quitting. Jack has lots of speeches about wanting to quit, but in a way, he doesn't actually act on. That's just cheap talk. Jack might, in fact, be [AUDIO OUT] if he were clean. But getting clean is too costly, because there are withdrawal costs.

Jack probably didn't [INAUDIBLE] [AUDIO OUT] tried cocaine, but this bad outcome was sufficiently unlikely that his early experiments with it made sense. So finally, cocaine should be legalized unless it generates externalities. That's essentially, people make choices.

They make-- if Jack started taking it, it must have been that he preferred cocaine over not taking cocaine, even though there's a chance that [AUDIO OUT] get really addicted and really be miserable now. And probably even if he keeps taking cocaine and [AUDIO OUT] to quit-- well, if he doesn't do it, well, then it must be a rational choice for him to, not because it must be that the withdrawal costs are really high.

Now we talked a little bit about this theory of rational addiction, much about Becker and Murphy. We talked about already that there's present bias or other factors involved, including biased beliefs. That might not be the case. [INAUDIBLE] understand that traditional economics, until very recently, would say that the rational thing to do with Jack here-- in fact, no good reason to make certain or any drugs illegal on externalities.

So externalities would be like if Jack causes behavior that's bad for others. Then well, we shouldn't let him do that, because other people are harmed and you might not internalize [INAUDIBLE]. But unless that's the case, we shouldn't intervene in any way.

OK, but so now, you might say, but what did we talk about in all [INAUDIBLE]? How is that consistent with what we talked about before? And the key of that question is the question of, do people act in their best interest? And so what we essentially assume often is that there is a rational relationship between people's choices, what they actually do, and the hedonic consequences of those choices, which are true well-being.

And so what we assume is essentially that you make certain choices that maximize your well-being. Essentially what you learned in 14.01 or 14.03 is essentially utility maximization, and that's at the heart of the rationality assumptions. So economists then tend to believe that most of the time, people act, at least approximately, in their best interest.

Now as we talked about already, and this is our previous lecture about Nisbett and Wilson telling more than we can know, we should be actually quite skeptical about the fact that people actually know what they're doing and why they're doing it, because [INAUDIBLE], as I showed you last time, in some sense, people often seem to have essentially no clue of why they do what they do. So it seems like a bit of an odd assumption to say, well, people are maximizing what's best for them.

So now how can we check whether this assumption is appropriate? What would you do? So I'm telling you here some behaviors. How can you actually check whether this assumption is an appropriate one? Lauren.

AUDIENCE: We could check their well-being after they make these decisions and if they're better or worse off.

FRANK SCHILBACH: Yeah, and we have to somehow figure out some situations where we can track people's well-being. Ideally, we would have some sort of experiment. There's group A and B, and in group A [INAUDIBLE], you let them choose, and in the other groups-- [INAUDIBLE] group C-- in some group, we essentially tell people, you should be doing A, doing one thing, and another group, we say, you should be doing another thing. And then there's another group where we have people essentially choose.

And then you can look at-- track people's well-being. And then we could figure out, is the group that gets to choose actually doing better than maybe some groups where we actually just tell them what to do?

And in some ways, or the work by Dan Ariely and others, if I can essentially manipulate you into making very drastically different choices by changing small things that really, arguably, should be irrelevant for your choice, like for example, the Social Security number-- for example, if I could manipulate you with anchoring you with your social security number and then say, would you like to purchase this item for $10, yes or no, and that affects your choice, well, then presumably, that's a clue in some ways that something is not going right. And so you are not necessarily optimizing, precisely because the Social Security number really should be entirely irrelevant, arguably.

Some forms of what we talked about, like preference reversal, or some form of demand from commitment or the like. If that's the case, it seems to be that then there's self control problems involved or biased beliefs in various ways that we all talked about, which seems then sent to say, well, you keep saying you want one thing. You keep spending money on, say, healthy eating, but then you never-- the food all goes bad in your fridge, and you keep eating donuts. That seems to be something that's not quite-- don't seem to be maximizing your utility.

But in general, it would be great if you could measure behavior and then the hedonic consequences of those behaviors, people's well-being, and if you then could say, well, option A or option B is better for people in general, or specifically when you ask them to choose, they can figure out what's best for them. Notice that you need to know the counterfactual here. You can't-- if I see you doing one thing and then track your well-being, it's not enough, because I don't know the counterfactual, whether you would be happy otherwise if you were not to choose that, OK.

So great idea. So now, how do we actually do that? So it's useful to have some terminology here that, in particular, that Kahneman and others have introduced. And Kahneman, remember, this is from Tversky and Kahneman, the Nobel Prize winner, who worked a lot on loss aversion and prospect theory that we talked about. Then Kahneman also has quite a bit of work on well-being and happiness and so on.

So and he introduces very useful terminology, which is-- they call decision utility and experience utility. So now what's decision utility? So economists tend to use the word utility or utility function to describe the preferences that rationalize observed choices.

Essentially, if you took 14.01, 14.03, et cetera, people talk about the utility function. Essentially, it's the function that you write down that maps your choices into well-being. And the function that you maximize, or by maximizing it, essentially say, well, given this utility function, given your behavior, it must be this utility function or that utility function allows me to rationalize what you're doing. That actually allows me then to explain why you choose apples over bananas.

And so Kahneman calls these revealed preferences the decision utility. That's essentially the preferences that rationalize decisions. When people make choices, that's the utility that they perceive or that they think they will perceive or receive as a consequence of their choices. That's essentially-- you might call that wanting or choosing.

So quickly, for an addict, the decision utility of drug consumption exceeds the decision utility of quitting. So if that person makes the decision of taking drugs, it must be that their decision utility is higher. When making those choices, they prefer one option over the other, which, again, doesn't necessarily then mean that afterwards, the experience is actually better. But when making the choice of taking the drug, they think that's a better option for them.

Now in contrast, Kahneman also talks about experience neutrality, which implies or talks about the hedonic consequences of choices. He calls these hedonic experiences experience utility. That's essentially the preferences that coincide with doing. So while you actually experience the whole consequences of your choice, what experience do you have, and what utility, or what's your well-being during that?

That's actually going back a long time ago. If you look at some of the wards in economics, Bentham and others, they talk a lot about pleasure and pain, which is very much about the experience people have in their life, as opposed to utility, which is, again, the construct that you have at the time of [INAUDIBLE]. Now how do we measure these hedonic experiences of well-being? So there's two questions, I guess.

One is, how can we measure this? If I wanted to know, what is your hedonic experience of this lecture right now, how would I do this? How do you enjoy, how much do you enjoy things? And then second, well, I can measure your experience right now, maybe this moment, maybe this hour, but somehow I need to aggregate those over time.

I need to know, you took class A versus class B. Now I can measure you at different points of time, but somehow, then I need to aggregate this overall experience. And how people are doing that is [INAUDIBLE] some questions.

Let's start with the first question, how do you measure people's well-being? If you wanted to learn about your classmates' experience in any type of situations, what would you do? What would you measure?

One is you can ask them verbally, like, how did you like activity A, how do you like B, or feel [AUDIO OUT] now? So this is asking what people want. But also just look at their facial features or the like and try to see how much they're smiling or the like.

Notice that that, you need to map into some happiness as well. In some sense, I could be smiling but be deeply unhappy, and then you would misinterpret that. So you'd have to have some function from facial feature to some underlying well-being that's somewhat grounded in some assumptions. It seems reasonable to think that when people are smiling, they're happy, and when they're crying, they're unhappy, but I'm just pointing out that that requires some underlying data to be able to map those things.

One is the physiological measures such as heart rates, [INAUDIBLE], et cetera. And so there, again, I think commonsensical stuff, like if you're really afraid and your heart goes up, presumably that's not a good thing. But if you're really excited and your heart rate goes up, presumably that's a good thing. So one has to add some assumptions or some structure on these measures. In some sense, you're just measuring them by themselves. In general, it might be misleading, but with some structure, [INAUDIBLE] use that well.

The other part is measuring-- there's lots of neuroscience and other work on this. Measuring people's brain and parts of the brain, and seeing which parts light up at which times could be another option. Of course, again there, you need to map that into something else. So you need to then be able to say, suppose I do something nice for you or [INAUDIBLE], and I can then figure out which parts of the brain are affected.

And then I can use that then to later, now if I wanted to figure out, suppose I could measure your brains during lecture. If I had figured out earlier which parts of the brain are associated with happy versus unhappy things, and now I could see what your brains are doing during lecture, I could then figure out whether you are happy or unhappy. But notice you need to have some mapping, because otherwise, it's hard to interpret what's going on. But of course, people have already worked a lot on that, that kind of mapping.

And just to be clear, so the revealed preference measures are-- hold on, let me just share what I have here. But the revealed preference measures are useful for some things, but in a way, if you wanted to measure whether revealed preferences give you a good answer, then of course, the real preference measure is not helpful. Or put differently, you want to check precisely. Suppose if I ask you, do you want lecture A or lecture B, and then everybody says they want lecture A-- or whatever, good A or good B, and everybody says they want good A-- the econ assumption is, by revealed preference, it must be that A makes you happier than B.

But if I wanted to the test, is that actually true, then I need to have your revealed preference, but then I also need to know your counterfactual in some sense. So I'd say give you good A, and I'd give you good B. How would that make you?

So I need the measures that we have here in the slides, a version of that, to say, OK, I'm going to look at your facial expression and see how happy you look when I give you good A versus good B. Or I look at your brain and see which parts of it-- do the pleasure parts of your brain light up? And if that's the case, I'm going to conclude that you're happier with good A versus good B, or if that's more so the case for good A.

And then I can look at your revealed preference. Is it really the case that you choose A over B, if the experienced utility says A makes you happier than B? Does that make sense? OK.

So what are these techniques? [AUDIO OUT] I think we covered most of [AUDIO OUT]. There's the observer ratings, facial [AUDIO OUT], reports of mood, pain, pleasure, happiness, and so on. This is what [AUDIO OUT]. There's autonomic measures, which is the respiratory, cardiovascular, your blood pressure, or your pulse, and so on, and so forth-- sweat.

There's vocal measures as well, which is pitch, loudness, tone. And you can probably detect whether somebody is angry or excited and so on. And again, you can apply some mapping here. You can look at people's brains.

There's also some responses to emotion sensitive tasks. So if I ask you, do you want to talk to a friend or do you want us to do something nice, and if the answer is, no, absolutely not, it could obviously be you're busy. But when people are in a bad mood, they tend to not want to do things that likely should make them happy.

OK, so we can try to do that. Now why might decision utility and experienced utility differ? That's of course what we discussed at length in the class already. There's various different explanations. One is-- we're going to talk about this a little bit, which is inaccurate memories of past experiences.

So you might just misremember your past experiences, and based on that, you might have poor forecasts of your future preferences or your utility. So this is essentially saying people's beliefs might be wrong in various ways, perhaps because people's memories are tricking them in some ways. People might also have wrong beliefs because they have some failures to anticipate adaptation, which is very much like projection bias that we talked about already previously. I'm going to talk about this a little bit more.

There could be things like emotions, which again, is in some ways often like projection bias, which is about things like people might be really angry, or people might be hungry. [AUDIO OUT] essentially it's very hard for them to make certain choices, failing to anticipate their anger or their hunger or other impulses [INAUDIBLE] over time. There's plenty funding of other reasons, but we talked about this quite a lot. And so you could think about a lot of what we discussed in the course about disconnects between decision utility and experienced utility.

Now I mentioned this already a little bit. The third part in people's utility is what Kahneman would call remembered utility, which essentially is to say, well, when you think about a past experience, you might not actually evaluate it correctly. In a sense, you might-- as I said, I might look at your experienced utility using these kinds of features, and then I might ask you afterwards, how did you like certain experiences? And then I can test, in a way, how are you aggregating, or how appropriately are you aggregating your experienced utility when I just ask you afterwards, how did you like certain experiences?

And one quite interesting feature in people's remembered utility is what's called-- excuse me-- duration neglect. People tend to remember the quality, rather than the length of the experience. And it seems to be also what's called the peak-end rule, which essentially says that retrospective evaluations are very much predicted by two things, or an average of the peak affective response recorded during an episode and the end value recorded just before the termination of the episode.

And how do I think about this? Well, mostly, I think this as being very salient. So when you think about and I ask you about a semester or a certain lecture or a certain experience that you had, maybe a TV show or the like, when you think about it and try to remember how much you enjoyed that experience, well, what's most salient, of course, is your latest experience. That's what comes to mind.

The last thing you remember is the last thing you experienced. And if the last episode of a TV show was terrible, that's what comes to mind. And you might say, well, the whole series was really bad, even though maybe previously, it was quite good.

And then the other thing, I guess, that seems to be really important is the peak, both positive or negative. So if something was really, really great or really, really bad, people really, really remember that, as opposed to stuff in between. And again, I think that's something that's just very salient in people's minds, and then in thinking about these experiences, that's what comes to mind.

And in studies that-- so what are these studies doing? Well, there are these studies that tend to measure experienced utility in certain ways, either through just measuring people's-- tracking people's well-being over time and then trying to predict their remembered utility, how did you like that experience. That's one version.

Another version is-- let me show you. Hang on one second-- is where people do certain tasks specifically, and they manipulate the experience deliberately. For example, there are these tasks which is called a cold pressor task, which is they ask people to put their hands into cold water and then vary the temperature, or they vary the duration. And then essentially, you can then try to predict people's remembered utility and then show things such as duration neglect or the like.

And let me show you what that looks like. For example, this is a famous study by Schreiber and Kahneman. They did a short trial, which is you put your hand into 14-degree water for 60 seconds, which is very cold, or a long trial where you put it into water for 60 second, and then the temperature rises to 15 degrees. I think this-- I hope this is Celsius, not Fahrenheit. But anyway, it's cold.

But the point here is that putting your hand for 60 seconds into 14-degree water, and then another 30 seconds into 15-degree water, these are both unpleasant experiences. So if you did each of these things in separation, people would say, [AUDIO OUT] just 60 seconds of cold water. I also don't like 30 seconds of 15-degree water. These are both unpleasant experiences.

So neither of them are actually desirable. If you ask people, if you give people [AUDIO OUT] do you want to experience this again, people say, hell no. I really don't like this, for both of these, in separation. But once you put them together, the 30 seconds of 15-degree water is more pleasant than the 14-degree water. So when you do that, people say they prefer the long trial over the short trial.

And the reason being is what they remember is the last experience, which was not that painful, when things got better over time. So people's remembered utility is violating any sort of rules of what you would usually think in terms of rationality and economics, which is precisely because they remember the end rather than the overall duration. So essentially, the utility does not aggregate in the way we would think they would, because the experience at each point in time [AUDIO OUT] negative.

There's more unpleasant studies for colonoscopy. Don't ask me about how they got ethical approval for those things. But anyway, a study [INAUDIBLE] where getting an extra minute where nothing bad happens. And essentially, the treatment group has better experience and memories of those kinds of-- of the overall experience. Now that sounds kind of like torturing people, but in fact, if you think about it, well, if-- in particular in medical procedures that are unpleasant overall, if we can help people have better memories of those experiences, maybe they will be a lot likely to come back and get future treatments or the like.

So in a way, you might say, what are these studies? Why are people torturing people? Why are researchers torturing people? I agree with that notion, but in fact, if you think about if you wanted to design, and if you think people are under-using certain medical procedures or certain medical tests that they should be doing-- think of dentists-- if there's some [AUDIO OUT] people's overall memory or memory of their overall experience better by having the end of that experience slightly better than the previous parts, that could [AUDIO OUT] helpful in generatings of behavior that's good for people in terms of their health, even though, of course, the last minute of the actual experience would not be pleasant.

Nobody wants to say, would you rather have another minute of colonoscopy? Absolutely not. But maybe the overall experience that you remember is, in fact, better than if not.

So there's other types of experience that shows this duration neglected and peak-end evaluations. People have shown pleasant and unpleasant movies or contents of movies. They have done these experiments with sounds or noise of various loudness and duration.

There's also animal [INAUDIBLE], where essentially animals are given unpleasant experiences, and then you look at what-- there's different types of experiences where the animal can go into one direction or the other, press one lever versus another, and then bad stuff happens in both of them, and one of them is worse than the other. And then you can look at, essentially, do animals have similar duration or other peak-end evaluations of experiences? And it seems to be very much a phenomenon that's also true for animals. Any questions on this?

OK. So then, of course, if that's the case, if people remember things differently, that will lead to violations of the decision utility maximizing the experienced utility. If you think about the experienced utility or the overall experienced utility of a certain choice that you make, in this case, I guess, the cold pressor task, think about this like the integral over, say, in this case, I guess, 90 seconds. In the first case, it's 60 seconds plus 30 seconds, 60 seconds feeling bad, versus the 30 seconds of putting your hand out, which is somewhat of a neutral experience.

If you're integrating that, presumably, the short trial would give you a higher valuation of higher, better, more positive experience overall. But then if you ask people, which one do you want, because people remember the long trial as being less bad, people might choose the longer trial over the short trial overall. So then that leads to, essentially, violations of the decision utility maximizing the experienced utility, because essentially, people remember things in a biased way.

OK. So then we talked about this very briefly. So so far, we talked about different ways of measuring. One specific thing that people have done a lot is just asking people simply about their happiness and life satisfaction. So simply, you ask people. There's different ways.

There are these types of ladder questions, how satisfied are you with your life as a whole these days? There's also affect questions about like, did you experience certain emotions yesterday? Did you feel angry? Did you feel happy? Did you feel sad? And so on.

The ladder question is a mixture of-- some people think of this as life satisfaction. Some people interpret it as happiness. Others would argue that it's rather about social comparisons.

So the ladder question has this notion of-- I'm going to show you the definition in a second, or in a bit. The ladder definition is very much just, on a ladder from 1 to 10, how great could your life be? Where are you in this ladder? So it has very much this flavor of some social comparison, which is not quite ideal, because it very well makes it like, well, on the ladder, how far and low are you at the ladder? It makes it very salient that some people are higher or lower than others.

So some of the responses that you see in those kinds of ladder questions are about happiness, but they potentially are also about people comparing each other to others in society and so on, which might not be necessarily the same as, how happy are you right now [INAUDIBLE]. Now what's problematic about these happiness questions, or why are economists skeptical, or what instances do we [AUDIO OUT]?

So just to be clear, so some people would argue, we should have a World Happiness Report and GDP. We should not maximize GDP or look at GDP growth as an overall performance of society. Instead, we should measure subjective well-being, and that's what we should essentially maximize. And policy should be-- the objective should be to maximize that. So what's the pushback on that?

So part of that is people and stuff that we talked already about, about remembered utility, or people just might not know what's good for them. But what's wrong with just asking people, how happy are you? And are people just going to tell you, or what's happening with that?

Yeah, so one problem, broadly speaking, is when you look at different countries, different societies, particularly, for example, if you look at Germans, Germans tend to be pretty grumpy and not super happy. So when you ask a German, how are you doing, the typical German would say, OK, by which they mean, actually pretty good. If you ask a typical American, how are you doing, they say, amazing. They probably mean, pretty good.

So when people give you certain statements, there's lots of cultural and other aspect that make comparison, particularly across societies, kind of tricky. You might think people in Latin America tend to report higher levels of happiness because that's a cultural thing, which might be that they're actually happier, or it might be that that's just the thing that you're supposed to exhibit. While people in some other states, Scandinavia, Germany, et cetera, might be, in fact, quite happy, but what they report to you is just some moderate level of enthusiasm overall.

So one is cultural comparison is quite hard using those types of questions. There's lots of Gallup surveys that people have done at large scales. It's the Gallup world happiness, or whatever, surveys. There's huge amounts of data that are being collected. They're a nationally representative surveys, and so on and so forth. So those data are available.

Of course, there's questions of, in some places, it's much harder to collect these than in others, and so on and so forth. And some people don't reply to surveys, and so on and so forth. For example, the elderly might be much less likely to answer some surveys, or some people without phones or internet, and so on and so forth, you might not be able to reach them. But in principle, you can solve all of those issues. But I think that it's just cost of use.

The reference dependence is a very interesting issue, because in some ways, you might say, well, now your income, but then your reference point goes off. Something happens to your reported happiness? But presumably you're happier than before? So that [AUDIO OUT].

On the other hand, utility might, in fact, be very much reference-dependent. For example, you might compare yourself to others. And then inequality, for example, might be extremely important for people's happiness. And one of the things that people argue is that, for example, [INAUDIBLE] the GDP, also for the US-- GDP in many places has gone up quite a bit, but happiness has not increased.

And people argue that's because, in part, inequality also went up a lot. And if then inequality really makes people deeply unhappy, well, that is important, and we should incorporate that in some ways in our considerations overall. But you see, that causes tricky issues, because in one case, doubling everybody's income, and presumably, people should be happier, because their living standard goes up.

So presumably, there, even if happiness doesn't go up, presumably, because living standards have gone up, and something is going wrong if I don't see increased happiness in terms of measurement-- on the other hand, there might be very much real issues about, inequality is quite important, and reference-dependent matters for people's happiness and well-being, in the sense that what really might matter a lot is relative income rather than absolute income, and we should capture that in some meaningful way. Oops. Sorry.

In particular, there's lots of issues with, depending on what I focus you on when I ask certain questions, I can get very, very different answers. And one, for example, example of this is a paper by Strack et. al. that looked at the correlation between general happiness and happiness with dating. And so what they did is a very simple thing. They essentially just manipulated the order of questions.

So first, they asked people about how happy are you in general. And then they asked people, how happy are you with your dating life? And in a different subsample, they asked people, how happy are you with the dating life, and then asked people about their general happiness. And what that does, essentially, it focuses people very much about dating as part of their life being important and for their happiness.

And similarly, I think this is exactly what Maya was saying as well. We need to ask, as you know, about the ladder question, which is about other questions about, depending on how I phrase the question, I can get very different answers and what you're saying. So that makes it very tricky to figure out which questions should we use, how much weight should be put on it. And even things like the order of questions matter. Matters quite a bit, which makes economists very nervous and so on, and then that's why, in part, economists would argue that we really should have revealed preference and choice.

Now on the other hand, I very much think the questions, when you just ask people and say, I'm deeply unhappy, that has some content, and there's information there. And so then the question is, well, how do we weigh these considerations? On the one hand, the survey instrument is really just very fragile, and we don't measure very much. On the other hand, there's useful information here that we should be incorporating somehow.

So I just wanted to have you-- that caveat in mind. And so let's now look at some data that people have collected. So what's a very nice source overall is what's called Our World in Data, which has lots of interesting graphs and figures where you can look at happiness and all sorts of things data on the world.

And when you look at life satisfaction around the globe, this is very much the ladder question here, which you see, exactly as I was saying. There's the ladder from zero to 10. And again, the ladder question is very much something hierarchical and makes inequality or relative comparisons quite salient. Even if it only asks you about your worst or your best possible life for yourself, it does have the feeling that really, it's about comparing yourself to others.

But anyway, that's very much used in many cases. Often people will call it life satisfaction, but even here, they also call it happiness overall. This is from the World Happiness Report.

And what you see is that, in general, rich countries tend to be-- report higher happiness, Scandinavian countries in particular. The US is reasonably happy. Europe is quite happy. And then if you go to poorer countries, in particular in Africa, they tend to report a lot lower happiness level.

When you look at comparisons across countries, you see a very clear relationship. Rich countries tend to be a lot-- report much higher life satisfaction compared to poor countries. That's a very clear association. That's true, both across countries but also true within countries.

So what this graph here shows you is the average income in a country. And then these arrows show you the gradient. Within the slope of the arrows is the gradient within countries. And what you see is essentially-- so this is comparing people within countries, rich versus poor countries. If you increase income within a country by $1,000 or something, how much does people's happiness go up?

And what you see is all of those countries have-- almost all countries have increasing gradients, which tells you that, both within and across countries, richer people report higher life satisfactions compared to poorer people. Similarly, that's also true for mental health. So these are all questions about subjective well-being, how happy are you overall, asking the overall population. But if you try to look at mental health conditions in terms of particularly depression and anxiety, so using clear definitions and psychiatric disorders for depression and anxiety and also for other conditions, you find that the poor in any given location are more likely to suffer from depression and/or anxiety.

Interestingly, the prevalence of depression, in fact, also and anxiety, is higher in rich countries. There's a lot of questions about measurement problems, like is it really measuring the same thing, and so on and so forth. But it could also be there's other factors involved, in particular, issues such as inequality. And it could just be that really, what's really important is relative income as opposed to absolute income.

Now we know also that anti-poverty programs improve mental health in various ways, including things like cash transfers and other anti-poverty programs. So this is an overview of studies. This is essentially a recent meta-analysis that we did for some [INAUDIBLE] paper of mine. And what you see here is a number of different interventions, both cash transfers and other anti-poverty programs, which are more broader programs. And what you see here is the treatment effect estimated in the different studies.

These are overall poverty anti-poverty programs. These are cash transfer programs. But overall, essentially, the overwhelming answer is that once you give people cash or reduce their poverty, that improves their mental health, as measured by different-- these are PWB, means psychological well-being indices. This is also depression screening scores and so on. So by various measures of people's mental health, giving people more money improves their mental health.

So that's true at the extreme in terms of really mental illness, depressive and anxiety disorders. It's also true for other measures of psychological well-being. If you just give people cash, they report higher happiness, in addition to what I told you before, this being true in the cross section. Now interestingly, we also tend to think that others are less happy than they say they are. So if we just ask people, how happy are other people in your country, almost every country in the world, what you find is that people tend to say the fraction that people actually say that they're very happy or rather happy is lower than they actually say they are, which is an interesting fact.

Now this-- as we talked about this already a little bit-- how is life satisfaction and happiness related to life events? In terms of when you look at different life events-- it's quite interesting that a lot of life events tend to affect people's reported well-being or life satisfaction a lot. For many events, including stuff like widowhood or divorce or marriage and other types of events, or winning the lottery, these effects are very much transitory, which is to say, when something really bad to somebody happens, their happiness goes down a lot. But it tends to recover quite a bit.

This is what we talked about before. People are quite adaptive to many changes, and they adjust quite well. This is what people call the psychological immune system, except for a few things, including unemployment. And one potential reason is that, for example, unemployment might also cause things like depression. And once you're really depressed, that doesn't really go back over [INAUDIBLE] that easily. That's true for women. It's also true for men overall.

There's also some other issues which you want to be careful of whether to interpret some of this data, which is there's called this evolution of latent situations, which is when you look at things like divorce, it tends to be that people-- when you look at people's divorce, you might say, well, divorce makes them happier in some ways. But really, what seems to be-- or you would look at this and say, marriage makes people less happy. But really, what seems to be the case is that people build up to a positive event, and then that peaks positive or negatively and then essentially goes back to what's before. For example, unemployment, things tend to go worse, even before people become unemployed, which seems to say, people tend to-- these underlying events, people tend to anticipate them already psychologically in some ways.

Now, this is what I was saying before. When you look at comparisons over time, you get sometimes these situations where, particularly in places including in China, but also in the US, where real income has risen a lot in many places, but yet people's life satisfaction does not tend to increase. Why is that? Well, that's, in part, I think, because of inequality going up, but [INAUDIBLE] inequality really staying the same and relative income really being important.

Could be other things going on. For example, pollution is really bad for people's health, mental health, but also their happiness and so on. It could also just be that people's reference points adjust, which is just, you make everybody twice as happy, and now everybody's-- the reference point goes up, even though their overall objective well-being standards, in terms of how much can they consume, et cetera, has gone up a lot.

Now this is quite interesting data now on ceiling effects. So there's some controversy over this. Some people are arguing that when you give people more money, at some point, there's a ceiling effect. It doesn't really make them happier anymore.

Now when you look at this graph here-- let me go back for a second. When you look at these graphs here, this-- whoops, sorry. [INAUDIBLE]. One second.

When you look at this graph here, it just doesn't look at all like ceiling effect. This looks like, essentially, self-reported life satisfaction as measured by these ladder question, essentially, is increasing, and there seems to be no ceiling, or it doesn't seem to be flattening in any way.

When you look at this one as well, it seems to be if you increase income by $1,000, at any level, it seems to be that reported life satisfaction goes up. So again, that doesn't look like a ceiling effect. And that's true as well if you look at these questions here.

Gallup had done these surveys in the US. When you look at the ladder question, which is this one here, when you look at annual income in the US, once you increase people's income, their mean ladder score, which is here on the axis on the right, tends to go up pretty much monotonically. Maybe it's a little bit flattening, but essentially, it goes up. So people report higher life satisfaction when they have higher income.

But when you look at people's affect, both things like positive affect, which is like happiness, smiling, enjoyment, not being blue, being stress-free and so on, what you see is-- in particular, this is for the US. When you go from something like $10,000 to $20,000 to $40,000 per year of income, there's a clear increase over time. So being poor is really-- or poor in the US, it is really, really bad for your affect and for your reported well-being overall.

But once you reach something like $50,000, $60,000, $70,000 in the US, it seems to me that there's no effect on positive affect, feeling blue, stress-free and so on. So one interpretation of that is that if we take these affect questions as more like a measure of experienced utility and happiness in any given life, essentially, having something like $70,000, $80,000 per year in the US-- granted, this is from 2009, so maybe this number has gone a little bit higher. But that number is sufficient to make people sufficiently happy, and beyond that, doubling of income will not do very much on how much you smile every day and how happy you are as a measure of those things.

In contrast, if I ask you the ladder question, which again, is very much about comparison, about how you feel yourself compared to somebody else in society, going from $100,000 to $200,000 will make, potentially, still quite a big difference. Any questions on that? OK.

So then let me-- this is among the most interesting parts of these kinds of studies, is the question about what, in fact-- so if not just income-- so if you know income essentially-- overall, if you're rich, you're going to be happier compared to if you're poor. But again, it seems to be really, if you want to not be unhappy, having income above $50,000 in the US is really important. Beyond that, it doesn't seem to do very much anymore.

But now what other things matter quite a bit? And this is a very simple exercise. So what they did is, this is [INAUDIBLE]. They looked at the Gallup survey questions from 450,000 Americans. And what they did is, for each of these four measures of psychological well-being, which is these here, positive affect, not blue, stress-free, and the ladder questions, they were first just trying to predict, what is the regression coefficient of high income? So if you just split the sample in two and look at high income versus low income, what is the regression coefficients of that?