This work has come out of my Undergraduate dissertation. I haven't shared or discussed these results much before putting this up.  Message me if you'd like the code :)

Edit: 16th April. After helpful comments, especially from Geoffrey, I now believe this method only identifies shifts in the happiness scale (not stretches). Have edited to make this clearer.

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 rescaling using long-run German panel data, looking at whether the association between reported happiness and three “get-me-out-of-here” actions (divorce, job resignation, and hospitalisation) changes over time.
• If people are getting happier (and rescaling is occuring) the probability of these actions should become less linked to reported LS — but they don’t.
• I find little evidence of rescaling. We should probably take self-reported happiness scores at face value.

1. Background: The Happiness Paradox

Humans today live longer, richer, and healthier lives in history — yet we seem no seem for it. Self-reported life satisfaction (LS), usually measured on a 0–10 scale, has remained remarkably flatover 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.

This is 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 agree on the empirical facts (i.e., self-reported happiness isn't increasing), 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” ?

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. 

I believe the method proposed below only identifies shifting in the happiness ruler: when the happiness of a 'best' and 'worst' possible life, both shift upwards. This process is illustrated below.

If rescaling is happening, then all the flat LS lines we see in national statistics might be wrong. It could mean that people are getting happier — we’re just measuring it badly.

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 that 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 work is interesting, but it totally relies on peoples' memories. We might employ motivated reasoning  to justify to ourselves that life is getting better. The authors attempt to correct for this

3. My Approach: Use Actions, Not Memories

If people are getting happier over time — but reporting it on a stricter, shifted scale — then the link between how happy someone says they are, and what they do when they're unhappy, should weaken over time. 

In other words: if life satisfaction is increasing, but the reporting scale is [edit] shifting, then big life decisions — like leaving a job or ending a relationship — should become less predictable from reported happiness.

This assumes that the underlying relationship between underlying happiness and 'exit actions' is: 1) negative and 2) convex (i.e., gets flatter at higher happiness levels). In other words, as peoples happiness with their family,increases to infinity, for example, the probability that they get divorced falls to 0.

Indeed, Kaiser & Oswald (2022) found that reported domain satisfaction was a strong, convex predictor of these GMEOH/exit actions:

If the scale has shifted upwards, we would effectively be sampling datapoints from further right on these graphs – where the relationship is flatter. This would result in a fall in the predictive power of life satisfaction (a flatter/weaker relationship).

I focus on three major exit or "get-me-out-of-here" (GMEOH) decisions:

• GMEOH_Marriage → Divorce
• GMEOH_Job → Voluntary resignation, transfer, or self-employment exit
• GMEOH_Health → Hospitalisation

and two measures of happiness, in the year proceeding this event:

• Domain satisfaction → Separate questions on satisfaction with family, work, and health (0–10 scale)
• Overall life satisfaction → “All things considered, how satisfied are you with your life?”

Method & Data

Using the German SOEP panel (1990–2022) — the longest-running panel dataset on happiness — I analyse:

• ~700,000 person-year observations
• ~6,500 divorces
• ~24,000 job exits
• ~83,000 hospitalisations

I run logistic regressions, predicting each GMEOH event using lagged happiness (life satisfaction in the previous period), and interact this with decade dummies to test whether the LS–action link weakens over time.

If the happiness scale is shifting upwards, then we’d expect the effect of LS on these actions to attenuate across decades.^[1].

4. Results

Exits Actions & Happiness with a Life Domain

• The relationships between life satisfaction in a particular domain – family, health, work – and a 'get-me-out-of-here' action were basically stable over time.
• No evidence of attenuation/rescaling here.

(I plot the basic models here – I also included individual fixed effects, but this made little difference and is more fiddly to plot.)

Exits Actions & Overall Happiness

I then combined all three exit actions into a single binary variable – which equals 1, if a person took any of the GMEOH actions in a year (became divorced, went to hospital, or quit their job). 

• I regressed this combined variable on overall life satisfaction (lagged) (including an interaction term for time periods)
• This increases statistical power, since domain-specific satisfaction isn't always measured.

Surprisingly, the relationship appears to get stronger over time. That’s the opposite of what rescaling predicts!

5. Conclusion

I started this project with a weak prior that maybe the Easterlin Paradox is a result of measurement error: maybe happiness really is rising — we’re just measuring it badly.

But I didn’t find that. Across three exit actions — divorce, job resignation, and hospital visits — the link with self-reported happiness is stable, or getting stronger. That suggests people use the happiness scale consistently over time. In short: rescaling (at least scale shifts) is not driving the Easterlin Paradox — at least not in Germany.

I also think the implications of the Easterlin Paradox are more profound than people realise. Decades of economic progress have made us no (or at least very little) happier! Isn't that insane!

More Stuff to do here

There's a bunch of limitations here, and a bunch more work that could be done. 

A few limitations that come to mind: 

• This identifies shifts, not stretches in the reporting function – I think!
• Divorce isn't voluntary – it isn't necessarily a 'get-me-out-of-here' action
• I could have done some non-parametric regressions – i..e., not logit. 

Other work

• This could be replicated on UK and/or Australian data. The Oswald/Kaiser paper also does this. 

1. ^{^}

   The results were robust to smaller time dummies too.