How robust is your assumption about the value life events staying constant? If it were not true, then there may not be any rescaling to explain. Intuitively, if wellbeing saturates at the top end, having a really positive thing happen to me genuinely might not move the needle as much. In other words, if my life is already a 9, is it realistic to expect getting married will take me to a 10—a perfect life?
HLI have a good, but very preliminary, look at the linearity/compression of wellbeing here, and it seems like linearity/compression is actually very under-studied. This seems odd to me, considering that it would probably dramatically shift where you allocate resources if linearity was true vs if there were bigger gains to be made in the middle of the spectrum.
I can't think of a good theoretical reason why true effects should fall so significantly – like 40%. That's striking. The same attenuation result holds, even including income/age/event prevalence.
"Intuitively, if wellbeing saturates at the top end, having a really positive thing happen to me genuinely might not move the needle as much."
This is true. Another way of saying this is: "the true effects fall as you get happier". But then, given reported happiness has stayed constant, why would the effects fall?
Hm, I don't think I agree with you on linearity. Andrew Oswald was writing about this in 2008. One option is that the function is logistic/arctan: i.e,. quite concave/flat at high latent happiness levels. That is, you can't shift reported happiness above a 10 (a ceiling effect), even if you get happier.
In this case: even if the reporting function is non-linear (and assuming true effect sizes are constant), why would the observed effects fall? Because people are getting happier. Again, this is a different way of saying rescaling is happening.
Re: assumption 1, "The underlying effect of life events is exactly the same", what if that's actually not the case? A couple brainstorming ideas on ways it's not
maybe some new environmental factor, like microplastics or hormone disruptors or something is changing the way we experience good and bad events, making them less salient?
maybe more hyper salient stuff like junk food, or emotional experiences from media like movies, is affecting how we experience those things?
(Idk how to indent on mobile) for example with movies, maybe vicariously experiencing an intense event, accompanied with a music score and everything else, leaves the real life event feeling dull in comparison? I've heard Sam harris touch on a similar point, where it used to be you only really got an up close, face-to-face experience with someone by actually being close to them, and you're "implicated" in it, your actions affect them and how they see you, whereas a movie you get part of the feeling of intimacy without being implicated, you can be slobbing it up on the couch and the result is the same
(Another indent) perhaps other forms of ragebait in the news and social media are more salient than life events, leaving actually frustrating things to have less of an impact?
afaik depression rates are increasing, maybe depressed people experience things less saliently? And we see effects of that across the spectrum even for "subclinical" depression?
Maybe if you know you'll be mostly ok even if a bad thing happens, whether from social safety nets or good planning or whatever, then it happening is less salient? Or for a good thing, being ok before it happens makes it less exciting, you're going from not-ok to ok, rather than ok to better
I'm sure there are others, but those are the main things I could think of. Not sure if they're true or not though
There's lots of points here. While they are possible, I would suggest they are not particularly common/well-suported in the psychological literature as it is today.
In addition, I don't know why these explanations would lead to desensitisation towards positive and negative events.
I wonder whether it would be better to measure subjective wellbeing by 1st finding the relationship between biological indicators (like heart rate) and self-reported happiness over a short period (like 1 day), and then tracking such indicators.
It would be interesting to know to which extent the reported LS predicts the corrected LS better than the logarithm of consumption.
Self-reported life satisfaction (LS), usually measured on a 0–10 scale, has remained remarkably flat over the last few decades, even in countries like Germany, the UK, China, and India that have experienced huge GDP growth.
I wonder which metrics assessing subjective wellbeing are more comparable across time. Self-reported happiness (which is different from LS) has increased with real gross domestic product (real GDP) per capita within the vast majority of countries.
Both are strong, difficult to verify assumptions. Briefly – on 1), I can't think of a good psychological theory why the true effects should be falling, after accounting for the variables mentioned above.
Have you accounted in some way for consumption tendentially increasing across time? People consuming more may be more or less resilient against life events.
By assumption 1): if the effect of life events has fallen by 40%, we know that the happiness scale might have stretched by a factor of 1/0.6 ≈ 60%
Nitpick. I think you mean the happiness scale stretched by 66.7 % (= 1/(1 - 0.4) - 1), which is roughly 70 %, not 60 %.
Calibrating with biological measures. Hm, could be a interesting, albeit labour intensive ... !
I've seen this graph a couple times on the Forum, now. I am confused why these lines are going up, but LS is generally flat. The one thing that stands out to me is that the timeframes are generally smaller than multidecade ones used for most studies on the Easterlin Paradox.
I'd also guess it'd be harder to calibrate the categorical response happiness question (This'd certainly be the case if you used my method, here.)
On income increasing over time. I discuss this more in the paper. We think that increasing income is the main pathway that rescaling occurs through. So, including it as a control could introduce over-control bias.
Oh, and I rounded from .62 something to .6 for the indexed effect size :)
What do you think of OWID's dissolution of the Easterlin paradox? In short:
OWID say Easterlin and other researchers relied on data from the US and Japan, but...
In Japan, life satisfaction questions in the ‘Life in Nation surveys’ changed over time; within comparable survey periods, the correlation is positive (graphic below visualises this for ~50 years of data from 1958-2007, cf. your multidecade remark)
In the US, growth has not benefitted the majority of people; income inequality has been rising in the last four decades, and the income and standard of living of the typical US citizen have not grown much in the last couple of decades
so there's no paradox to explain.
(I vaguely recall having asked you this before and you answering but may be confabulating; if that's happened and you feel annoyed I'm asking again, feel free to ignore)
I don't think you did mention this before...! I think this graph is just for 1 country. Perhaps Japan.
To be honest, I don't know what to think of the Wolfers/Stevenson objections! My only thought is: differences of, e.g,. 0.2 points, would look pretty small in comparison to the potential rescaling effects I suggest here.
self-reported happiness over a short period (like 1 day)
Not exactly what you meant, but you may be interested in Jeff Kaufman's notes on his year-long happiness logging self-experiment. My main takeaway was to be mildly more bearish of happiness logging than when I first came across the idea, based on his
Overall my experience with logging has made me put less trust in "how happy are you right now" surveys of happiness. Aside from the practical issues like logging unexpected night wake-time, I mostly don't feel like the numbers I'm recording are very meaningful. I would rather spend more time in situations I label higher than lower on average, so there is some signal there, but I don't actually have the introspection to accurately report to myself how I'm feeling.
Scattered quotes that made me go "huh":
When I first started rating my happiness on a 1-10 scale I didn't feel like I was very good at it. At the time I thought I might get better with practice, but I think I'm actually getting worse at it. Instead of really thinking "how do I feel right now?" it's really hard not to just think "in past situations like this I've put down '6' so I should put down '6' now".
Being honest to myself like this can also make me less happy. Normally if I'm negative about something I try not to dwell on it. I don't think about it, and soon I'm thinking about other things and not so negative. Logging that I'm unhappy makes me own up to being unhappy, which I think doesn't help. Though it's hard to know because any other sort of measurement would seem to have the same problem.
Thanks for sharing, Mo! Very interesting. That makes me more pessimistic about finding the relationship between biological indicators and self-reported human welfare. I still think tracking more objective metrics would be helpful such that is is harder to game the system. If welfare surveys became widespread, and consistently used to make decisions, people could try to give answers which benefit them the most instead of reporting their welfare as accurately as possible. I like the assumption that welfare per human-year is proportional to the logarithm of annual consumption because this is hard to game.
This is such a cool and interesting post. And it totally helps understand why things for humans seem like they're materially getting better in many ways, but our culture involves a lot of people saying a lot of negative things about how humanity is doing. I love that you're challenging the assumption that people accurately remember, or accurately report their memory of, their own happiness! This is super smart and matches my own experience of what it's like to be a human. Often, I'm not super sure how happy I've been over the past week, or the past day. When someone asks me, it feels like a hard question. All I can remember is how I feel right now. And even that can sometimes be hard to articulate.
And thank you for mentioning that GDP growth could still be bad due to its effects on animals!
Self-report, is like popular sentiment unreliable. If you'd use Schopenhauer's method, you'd measure happiness by lack of privation. Remove a thing, see if the person feels worse.
Using numbers seems to have a limit.
"As remained remarkably flat over the last few decades, even in countries like Germany, the UK, China, and India that have experienced huge GDP growth."
Sleep deprivation, screen addiction, and increase of sight problems are a likely cause, urban sprawl and so on...
A high GDP's not the problem...
Are kids ever polled for happiness? If you take only one sample, it's gonna be bad. If you take a smaple every year, from 3 to 70, it'll work better.
Easterlin paradox isn't true.
If you want to measure, in a less skewed way, use MORE indicators, avoid Hoefstader's law.
It's also false, that growth did not benefit the average person, since it did. How many had e-readers, wireless headphones, and flat-screens in 2005?
Maybe, just age makes people sad? Maybe
That makes me more pessimistic about finding the relationship between biological indicators and self-reported human welfare.
Weight? Reported hours of sleep? Nearsightedness out of 10?
Around 1 month ago, I wrote a similar Forum post on the Easterlin Paradox. I decided to take it down because: 1) after useful comments, the method looked a little half-baked; 2) I got in touch with two academics – Profs. Caspar Kaiser and Andrew Oswald – and we are now working on a paper together using a related method.
That blog post actually came to the opposite conclusion, but, as mentioned, I don't think the method was fully thought through.
I'm a little more confident about this work. It essentially summarises my Undergraduate dissertation. You can read a full version here. I'm hoping to publish this somewhere, over the Summer. So all feedback is welcome.
TLDR
Life satisfaction (LS) appears flat over time, despite massive economic growth — the “Easterlin Paradox.”
Some argue that happiness is rising, but we’re reporting it more conservatively — a phenomenon called rescaling.
I test this hypothesis using a large (panel) dataset by asking a simple question: has the emotional impact of life events — e.g., unemployment, new relationships — weakened over time? If happiness scales have stretched, life events should “move the needle” less now than in the past.
That’s exactly what I find: on average, the effect of the average life event on reported happiness has fallen by around 40%.
This result is surprisingly robust to various model specifications. It suggests rescaling is a real phenomenon, and that (under 2 strong assumptions), underlying happiness may be 60% higher than reported happiness.
There are some interesting EA-relevant implications for the merits of material abundance, and the limits to subjective wellbeing data.
1. Background: A Happiness Paradox
Here is a claim that I suspect most EAs would agree with: humans today live longer, richer, and healthier lives than any point in history. Yet we seem no happier for it. Self-reported life satisfaction (LS), usually measured on a 0–10 scale, has remained remarkably flat over the last few decades, even in countries like Germany, the UK, China, and India that have experienced huge GDP growth. As Michael Plant has written, the empirical evidence for this is fairly strong.
(I note this empirical literature is contested. Often it boils down to small differences – has happiness increased from a 7 to a 7.2? My results suggest these differences may be negligible, given potential rescaling effects.)
Stagnant average happiness is known as the Easterlin Paradox. It is a paradox, because at a point in time, income is strongly linked to happiness, as I've written on the forum before. This should feel uncomfortable for anyone who believes that economic progress should make lives better — including (me) and others in the EA/Progress Studies worlds.
Assuming we agree that self-reported happiness is (close to) flat, there are a few potential explanations:
Hedonic adaptation: as life gets better, our expectations rise just as fast — so we don’t feel happier.
Social comparison: we care about relative, not absolute, gains.
Rescaling: maybe happiness is increasing, but the 0–10 reporting scale has shifted or stretched.
It’s that third one — rescaling — that I try to empirically test here.
2. What is “Rescaling”?
(AKA: 'scale norming', 'scale shifts')
The rescaling hypothesis suggests that the reporting function — how we map our true wellbeing onto a 0–10 scale — changes over time. You can think of the reporting function like a ruler, converting underlying happiness to reported Life Satisfaction. Suppose two people, one from 1990 and another from 2020, both report their Life Satisfaction (LS) = 8/10. The second person could be happier, if their reporting function (i.e., their ruler) has either shifted or stretched upwards.
The method proposed below only identifies stretches in the happiness ruler: when the quality of the average "10/10" life improves, but a "0/10" life stays roughly the same in underlying happiness terms.
Here's a visual illustration of how this could bias national happiness data:
If the LS scale is stretching upwards, we could see constant self-reported LS, alongside improving underlying happiness.
There’s very little work testing this directly: probably less than 10 papers, and only 1 relating to the Easterlin Paradox (Prati & Senik, 2025). They use a memory-based approach: ask people how happy they used to be and compare to what they said at the time. If I said “7/10” in 2010, but remember it as “6/10” in 2015, that suggests my internal scale has changed. Using this method, Prati & Senik estimate that happiness in the U.S. might be underreported by 80–140% (!).
This method – using recalled happiness to adjust for rescaling – was discussed on the EAF recently. It runs into two issue: 1) what if we misremember our happiness in the past? 2) what if we correctly remember our previous happiness, but underreport this due to motivated reasoning/social desirability (not rescaling)?
Using memories to demonstrate rescaling is interesting, but I think we need more literature to be confident that rescaling is a real problem.
3. My Approach: Life Events would look smaller on stretched out rulers
I propose a new way to test for rescaling: track how people respond to major life events over time. The logic is simple. If happiness scales are stretching, the same event (say, unemployment) should appear to have a smaller emotional impact in later years. A way of conceptualising this is:
If a ruler has been stretched out, if we were to measure a given object, it would look artificially smaller.
The picture below illustrates this:
This has the assumption that life events should correspond to a similar amount of underlying happiness over time. (Later, I attempt to control for structural reasons why this might change.)
Using the German SOEP panel (1991–2022, ~650k observations), I focus on six major life events: four negative (partner’s death, death of a parent, unemployment, separation) and two positive (new relationship, cohabitation). I selected these from a candidate list of 13 events, based on data availability and whether the reported effects were unambiguous in every period.
Then, I looked at the relationship between whether a life event occured in the previous year, and the change in an individual's reported LS (taking the difference accounts for unobserved things about each person).
4. Results: Effects Are Shrinking
The absolute effects of 3 life events (unemployment, cohabitation, death of parent) were converging to zero, but for the remaining 3 events (death of partner, separation, new partner) the effects were relatively stable. This constitutes some non-definitive evidence for rescaling – which would predict, all else equal, attenuation for all events.
However, first, some events had much smaller standard errors than others; and second, detecting attenuation is pretty hard for such small effect sizes (typically 0.2-0.5 on an 11 point scale!). To smooth out noise, I conducted a meta-average: setting all effect sizes to 1 in the first period, and giving more weight to events with smaller standard errors (i.e., estimates were are more certain about: this is because, for example, unemployment is more common than bereavement).
The meta-analysis indicated that the average effect of a life event had fallen by ~40% over three decades (1991-2022). The average effect of a life event is monotonically converging towards zero. Interesting.
In the paper I attempt to show this result is robust to a bunch of different specifications. If you look at the Google Doc linked above, you'll see a lot of graphs like the one shown above – with the average effect of a life event trending towards zero. This occurs regardless of:
Whether we control for demographic factors (e.g., changes in education, age, gender make-up in the population) which affect happiness
Whether we allow the effect of life events to vary by age or the prevailing frequency of life events in the population
Whether we look at all occurrences of life events, or just the first instance (e.g., your first time being unemployed, or any time)
Whether we look at change over the course of four 8-year periods (or 2 x 15-years, 6 x 5-years, or 10 x 3-years).
If we assume the real psychological impact of these events has been roughly constant, then the reporting scale must have stretched. In other words, there is decent evidence that rescaling is occurring.
5. How much might we be underestimating life satisfaction?
Here, I make 2 assumptions:
The underlying effect of life events is exactly the same
The happiness of the average "0/10" life has remained exactly the same
Both are strong, difficult to verify assumptions. Briefly – on 1), I can't think of a good psychological theory why the true effects should be falling, after accounting for the variables mentioned above. If you can think of one, I'd love to know. On 2), I note, happiness researchers tacitly make this assumption anyway! Violations of either assumption could mean that my BOTEC adjustment is a over- or under-correction for rescaling.
(I also assume that the reporting function is linear/evenly spaced. There is decent evidence for this in the literature).
By assumption 1): if the effect of life events has fallen by 40%, we know that the happiness scale might have stretched by a factor of 1/0.6 ≈ 60%
By assumption 2): we can adjust for rescaling by multiplying average reported happiness by 1.6.
I do this process for every period, and plot adjusted LS alongside reported LS and GDP (and the meta-average index). Interesting, I arrive at a fairly ball-park number, of how much happiness might be underestimated (60%) to Prati & Senik (80-140%).
This offers a possible explanation for the Easterlin Paradox: maybe people are getting happier — actually quite a lot happier – but, our reporting standards have changed.
6. Implications
I feel like there are two main implications of this. But before I discuss them, I'd like to say: the upshot here is still tentative.
I think the evidence that rescaling is happening is decent. However, we're working with pretty uncertain priors here, given there is so little literature on rescaling. The evidence that people are much happier than in the past – even more tentative. Though, to suggest that average national happiness has remained the same, we'd need to suggest that a lot of downward shifting has occurred (the quality of a "0/10" life has worsened a lot). This is possible, but strikes me as unlikely at the moment, given the literature on rescaling assumes that shifts occur upwards (the quality of a "10/10" life has, if anything, improved).
Two implications of this work:
1) GDP might be much better for human happiness than we previously thought
I note, GDP growth could still be, on net, bad (or very bad), given its effects on animals (and/or x-risk).
2) There might be a large amount of bias in nominal LS data.
I don't know yet, how rescaling might bias happiness-based cost-effectiveness work. I'd presume it could matter quite a lot. My method is much too coarse to calibrate nominal happiness for a given person or even for a given country within a particular year, so would not help. Alternative methods for adjusting for rescaling – e.g., memory – have their own problems (see Section 2).
More work here might be interesting. Please message me if you'd like to chat about this stuff more, or if you'd like my code.
AI Use Note: Main body text entirely human written. Claude (Opus 4.8) helped develop models of animal life histories in the appendix.
Cross-posted from Good Structures.
Executive Summary
* Animal advocates sometimes make claims like “there are X of this animal...
“How long have you been v*g*n?”
This is one of the most common icebreakers at animal protection events. It’s a baseline assumption, and it mostly holds true: if you’re out advocating for animals not to be tortured or abused, realistically these days you are v**n, or close. And it makes for good conversation. It seems fairly safe to assume when you meet strangers.
But this assumption is hurting the movement in a way which we don’t always notice: someone new comes into the sp...
Summary
Back in November 2023 I posted here to launch Spiro and raise our first $198k. Two and a half years later this is an update and a fundraiser for the next step.
The short version: we've now reached over-5,900 people with TB preventive medicine, including over 3,000 children under five years old. Our early results have held up well an...
How robust is your assumption about the value life events staying constant? If it were not true, then there may not be any rescaling to explain. Intuitively, if wellbeing saturates at the top end, having a really positive thing happen to me genuinely might not move the needle as much. In other words, if my life is already a 9, is it realistic to expect getting married will take me to a 10—a perfect life?
HLI have a good, but very preliminary, look at the linearity/compression of wellbeing here, and it seems like linearity/compression is actually very under-studied. This seems odd to me, considering that it would probably dramatically shift where you allocate resources if linearity was true vs if there were bigger gains to be made in the middle of the spectrum.
(Apologies if you have addressed this somewhere)
Hello, Huw!
I can't think of a good theoretical reason why true effects should fall so significantly – like 40%. That's striking. The same attenuation result holds, even including income/age/event prevalence.
"Intuitively, if wellbeing saturates at the top end, having a really positive thing happen to me genuinely might not move the needle as much."
This is true. Another way of saying this is: "the true effects fall as you get happier". But then, given reported happiness has stayed constant, why would the effects fall?
Hm, I don't think I agree with you on linearity. Andrew Oswald was writing about this in 2008. One option is that the function is logistic/arctan: i.e,. quite concave/flat at high latent happiness levels. That is, you can't shift reported happiness above a 10 (a ceiling effect), even if you get happier.
In this case: even if the reporting function is non-linear (and assuming true effect sizes are constant), why would the observed effects fall? Because people are getting happier. Again, this is a different way of saying rescaling is happening.