That single pooled number would hide the most interesting thing in the data. Fit the relationship once, pooled across the world, and you blend 155 cultures, confound rich and poor nations, and let a few giant countries set the tone. Fit it separately inside each country and one coefficient becomes a distribution of 155 of them. The spread is the story.
Evaluative — the life rating
“Please imagine a ladder, with steps numbered 0 at the bottom to 10 at the top. On which step do you stand?”
ladder · 0–10 · a reflective judgment
Experiential — the mood
“Did you experience enjoyment / worry / sadness during a lot of the day yesterday?”
pos_affect, neg_affect · 0–1 shares · felt, not judged
These two are not the same thing, and a half-century of research insists we keep them apart. A life rating is a verdict you reach when asked to take stock; a mood is what actually coloured your yesterday. Kahneman and Deaton (2010) found that in American data, emotional well-being flattened out around a household income near $75,000 while the life rating kept climbing. The argument over whether that plateau is real ran for a decade — Killingsworth (2021) found no plateau in experienced feeling, and the three principals eventually settled it in an adversarial collaboration (2023). A sibling piece in this series, plateau in the tail, re-runs that fight in US data. This is the global companion: not whether the slope bends, but how the two slopes compare inside every country at once.
1Two distributions, one asymmetry
The method is plain. Within each country, weighted by Gallup’s within-country weight, regress each outcome on the respondent’s income quintile (1 to 5, defined inside that country, which sidesteps the impossible task of comparing dollars across borders) plus a standard set of controls — age, age squared, sex, education, marital status, and household size. Keep the one coefficient on income. Do it 155 times.
The raw slopes already tell the first half of the story. The median country gains about 0.27 ladder points per income-quintile step, and the slope is positive in every single country that clears the bar. Positive affect rises with income too — a median of +0.019 on its 0–1 scale, up in 100% of countries — and negative affect falls in 98.7%. Money, within nations, reliably buys both a better verdict and a better day.
But the ladder runs 0 to 10 and affect runs 0 to 1, so the raw slopes can’t be compared directly. To put them on one scale, each outcome is z-scored within its own country before refitting — expressing every slope in within-country standard deviations per quintile step, which also nets out the cross-cultural habit of using answer scales differently. Now they line up. And they do not match.
The life-rating slope sits to the right of the mood slope
Each tick is one country’s standardized income slope (within-country SDs per quintile step). Top strip: the ladder. Bottom strip: a single mood summary combining more enjoyment and less worry/sadness. Diamonds mark the population-weighted “representative person.”
Standardized, the asymmetry is sharp: in 96% of countries the life-rating slope is steeper than the mood slope. The evaluation premium — the standardized ladder slope minus the standardized mood slope — is positive in the same 96% of countries, with a median of about 0.06 standard deviations. Read plainly: an extra income quintile lifts how people rate their lives noticeably more than it lifts how they feel. The verdict is more sensitive to money than the mood is, almost everywhere.
The same income step buys more life rating than mood in 96% of countries on Earth.
None of this means money transforms well-being. A quintile step is a quarter of a ladder point and a couple of percentage points of daily feeling — modest, durable nudges, not metamorphosis. The point is the relative tilt: across the income gradient, the reflective verdict bends more than the felt experience.
2Does the premium grow with wealth?
If the verdict is more money-sensitive than the mood, a natural next question is whether that gap is bigger in richer countries — where, the story goes, status and comparison do more work on how people rate themselves. The spread of the premium across 155 countries is itself data. Regress each country’s premium on a national moderator, weighting countries by the inverse of their estimate’s variance (the standard meta-analytic weight), and the slope of the slopes tests that hypothesis directly.
The evaluation premium against national development
One dot per country, sized by population. Vertical axis: the standardized evaluation premium Δ. Toggle the horizontal axis between three country attributes; the line is the inverse-variance-weighted fit.
Measured by the Human Development Index, the premium does rise with development: the meta-regression slope is positive and its confidence interval clears zero (slope ≈ 0.055, R² ≈ 0.08 across 152 countries). Move from a low-HDI country to a high-HDI one and the gap between the life-rating slope and the mood slope visibly widens. The richest, most-developed societies are where the verdict pulls furthest ahead of the feeling.
Honesty requires the caveat. Swap HDI for raw GNI per capita and the gradient nearly vanishes — the slope is positive but tiny and barely distinguishable from zero (R² ≈ 0.03). Development as a broad bundle — health, schooling, and income together, which is what HDI captures — tracks the premium; income alone does much less. The composite, not the cash, is what moves the gap.
3Is the income tilt steeper where inequality is higher?
A second, separate hypothesis: maybe income matters more to your life rating where the income distribution is more unequal — where the rungs of the ladder are further apart. To test it, regress the raw life-rating slope itself on each country’s Gini index. Here the data decline to cooperate. The meta-regression slope is essentially flat — slightly negative, with a confidence interval straddling zero and an R² near 0.01. Within-country inequality, at least as the Gini measures it, does not detectably steepen the income–ladder gradient. That is a null, and it stays in the piece: the income tilt is remarkably similar whether a country is flat or sharply stratified.
4The geography of the gap
Mapping the life-rating slope shows how even it is. The income–ladder gradient is positive on every continent; what varies is its steepness, not its sign. The premium — where the verdict outruns the mood — is where the cross-country structure lives.
Where money tilts well-being, country by country
Diverging scale centred at zero. Toggle between the raw life-rating slope and the standardized evaluation premium. Grey countries are suppressed, unmatched in this atlas, or absent from the poll.
The steepest income–ladder slopes cluster where the climb out of hardship is most consequential — sit at the top. The premium — the verdict pulling ahead of the mood — runs highest in places like , and is smallest, even faintly negative, in a handful of low-income countries such as , where money moves the daily mood about as much as it moves the rating.
5What this does and doesn’t show
The robust finding is the asymmetry and its sign: in 96% of countries the income gradient is steeper for the life rating than for the mood, and the premium widens with broad development. The effects themselves are modest — about 0.27 ladder points per income quintile, not a transformation — and small cross-country differences in slope are not, on their own, definitive, because cultures use answer scales differently. The within-country z-scoring is designed to blunt exactly that response-style problem, and the asymmetry survives it.
Two robustness checks sit behind the drawers below. Adding employment status as a control (only recorded from 2009 onward, which trims the panel) leaves the premium essentially unchanged. Swapping income quintiles for the logarithm of imputed per-capita income — a noisier, heavily imputed measure — keeps the ladder slope positive everywhere. The headline does not depend on either choice.
Robustness — add employment status (2009 onward)
Robustness — log per-capita income instead of quintiles
Meta-regression weighting — inverse-variance vs. population vs. unweighted
Show the full country table (155 countries)
| Country | n | waves | β ladder | z ladder | z mood | premium Δ |
|---|
What the data can’t do is settle why. That the verdict is more money-sensitive than the mood is consistent with several readings — that life ratings absorb social comparison and status, that they pull in expectations about the future, that felt experience is simply harder to buy — and this design can’t adjudicate among them. The static country attributes are a single ~2015–2018 snapshot joined to every year, so the development gradient is cross-sectional, not a story about change over time. And “population” here is total population standing in for the adult population Gallup actually samples.
The novel piece is the framing: the evaluative-minus-experiential slope gap as a country-level quantity, mapped across the full 2005–2020 panel, and its gradient with development. Earlier work — Jebb and colleagues (2018) on income satiation by region in this very dataset, Stevenson and Wolfers (2013) arguing there is no satiation at all, Diener, Tay and Oishi (2013) on rising income and well-being — mostly studied the shape of one slope. Here the two slopes are watched together, in every country, and the gap between them turns out to be the durable signal.