The whole book has run on one image: a pair of scissors. The upper blade is how people rate their lives — the considered, comparative verdict, resilient and slow to bend. The lower blade is how people feel their days — the worry, the enjoyment, the sense of someone there — which has thinned and, since 2020, has not come back. The gap between the blades was the story. This coda asks the practical version of it: if you wanted to lift either blade, what would you reach for first?
Three datasets answer in three different vocabularies, so the honest move is to keep them in separate rooms. In each one, every predictor is standardized within its own country and year, then run against a single well-being outcome — first alone, then against all the others at once. The number to watch is the standardized coefficient: how much the outcome shifts for a one-standard-deviation move in that predictor, holding the context fixed. The conventional wisdom going in is that income is the prime mover. It is not what the data says.
Two blades, named again
This book never lets the two kinds of well-being share an axis, and the coda is no exception. Evaluative A standing judgment of life as a whole: the GSS three-point happy question, Gallup's 0–10 ladder, and the World Values Survey's 1–10 life-satisfaction scale. Experiential What yesterday actually felt like: Gallup's binary positive- and negative-affect items. The split matters here because the surveys are lopsided. Only the Gallup World Poll carries experiential items at all; the GSS and the World Values Survey have no daily-affect questions, so the lower blade can be measured in exactly one of the three rooms. Where it can, the contrast is the punchline.
The horse race, one dataset at a time
Standardized OLS coefficients on each dataset’s evaluative well-being outcome. Filled dot = the predictor run alone (bivariate); hollow dot = the same predictor net of all the others (joint). The whisker is ±1.96 standard errors. Freedom and connection are drawn in gold against income’s steel reference line. Switch between running each predictor alone and running them all together.
wtssps; Gallup World by within-country wgt; WVS by within-country weight. Each predictor is z-scored within country×year (year alone for GSS). Health is absent from the Gallup panel and freedom from both Gallup and GSS — the surveys simply do not ask — so those rows are blank, not zero. WVS substitutes felt financial satisfaction where it has no social-support item.Income’s power is mostly borrowed
Run income alone and it looks like the answer everyone expects. In the World Values Survey, a one-standard-deviation step up the subjective income ladder is worth +0.48 points of life satisfaction; in the Gallup World Poll it moves the Cantril ladder +0.38. Respectable numbers. Then put income in the room with health, freedom, marriage, work and faith, and most of that pull turns out to have been on loan. In the WVS, income’s coefficient collapses from +0.48 to +0.06 once the other predictors are present — richer people are also healthier, freer and more partnered, and when you hold those constant the money has little left to do on its own. The GSS tells the same story in miniature: income’s bivariate +0.11 on happiness falls to +0.04 net of everything else.
What survives the joint model is more interesting than what doesn’t. In the WVS, Evaluative perceived freedom and control over one’s life is the second-strongest predictor in the whole field — +0.79 alone, +0.48 net — beaten only by how satisfied people feel with their household finances, which is a feeling about money, not money itself. Self-rated health holds at +0.32 net. Income, the thing the public imagination reaches for first, finishes near the back of the pack.
Connection ties income, then beats it
The Gallup World Poll cannot ask about health or freedom in this extract, but it can ask the question this book cares about most: do you have someone to count on? Run alone, that single yes/no item moves the ladder +0.37 — statistically a dead heat with income’s +0.38. Put both in the joint model and the order quietly flips: social support lands at +0.34 and income at +0.34, with connection a hair ahead. Two of the most reliable levers on how people rate their lives, in the largest well-being survey on Earth, are a person to lean on and a sense that the life is your own to steer. Neither is money, and one of them is free.
The cross-blade test — Gallup World Poll only
The same predictors, now on both blades. Left column: the 0–10 ladder Evaluative. Right columns: positive and negative daily affect Experiential, both oriented so up is good (the worry/sadness index is sign-flipped for display). Bivariate coefficients. Watch income shrink as you cross from the upper blade to the lower one.
wgt, 2006–2025; predictors z-scored within country×year. The ladder is on a 0–10 scale; the affect indices are 0–1 shares, so the columns are on different rulers — read each column on its own axis. Negative affect shown as “absence of negative affect” (sign-flipped) so that higher always means better.This is the cleanest single fact in the coda. Income moves the upper blade hard and the lower blade barely. On the ladder, a standard-deviation of income is worth +0.38; on freedom from daily negative feeling it is worth +0.035 — roughly a tenth as much. Money buys a better rating of your life far more readily than it buys a better day. Social support, by contrast, holds its size across the divide: it is the top mover on positive affect (+0.046) and on the absence of negative affect (+0.042), and it ties income on the ladder. Connection is the rare lever that works on both blades at once. That is the whole argument of the book, recovered from a regression instead of a trend line.
Where the surveys can’t see
Three datasets that each lack what another has is not a tidy design; it is the honest one, and pretending otherwise would undo the book’s own discipline. The freedom result rests on the World Values Survey alone, because neither Gallup nor the GSS asks about it. Health is missing from the Gallup panel for the same reason. The GSS and WVS carry no experiential items at all, so the cross-blade test is a Gallup-only result. These are not modeling choices to be optimized away — they are gaps in what humanity has bothered to measure, and the chart leaves them blank rather than filling them with a confident zero.
Gallup World Poll
No self-rated health item and no freedom-of-choice item in this extract. Its strength is scale (164 countries) and the only experiential affect measures in the three.
World Values Survey
No comparable “someone to count on” item, so connection can’t enter; felt financial satisfaction stands in as an income-adjacent proxy. No daily-affect items.
GSS
No freedom/autonomy item and no experiential affect items across its whole run. Its gift is the half-century of time it covers.
The predictors that won’t sit still
Not everything points the same way across rooms, and the disagreements are worth naming. Marriage is the clearest: in the GSS it is one of the stronger predictors of happiness (+0.15 alone, +0.10 net), and in the World Values Survey it reads +0.15 — but in the Gallup World Poll, against the ladder, married people sit very slightly lower (−0.02) once you standardize within country. The sign genuinely flips. Whether that reflects who marries where, what marriage means across 164 countries, or the company marriage keeps in each model, three surveys cannot adjudicate. Employment and religious attendance are similarly modest and context-bound — positive and small in most rooms, occasionally negative in the joint models. The reliable through-line is the one the book has been making all along: the levers that move a life most are health, the felt freedom to run it, and the people in it — not the size of the paycheck.
What this settles, and what it doesn’t
It is a horse race, not a causal verdict. These are weighted associations within country and year, not experiments. Freer, healthier, better-connected people may rate their lives higher for reasons the regression can’t see, and the arrow could run the other way — a good life makes a person feel free. The coda shows which doors are worth knocking on, not that knocking guarantees an answer.
The scales don’t line up, on purpose. A coefficient on a 0–10 ladder is not a coefficient on a 3-point happiness scale; the predictors are standardized but the outcomes keep their native rulers, which is exactly why the three panels stay apart. The cross-dataset claim — that income looks far weaker for the lower blade than the upper — lives inside the Gallup panel, where the comparison is fair, not across the seams.
With those limits stated plainly, the book’s closing bet comes in. For decades the public conversation about well-being has been a conversation about money. Run the contest in three of the largest surveys ever fielded, and money is rarely the winner and never the runaway. Health rivals it. Perceived freedom beats it. A person to count on ties it on the upper blade and laps it on the lower one. The scissors opened because the experiential blade fell while the evaluative blade held. What this coda adds is the repair manual: the things that would close the gap are not, mostly, the things we spend our public life chasing.
Notes & data
- Sources. GSS 1972–2024 cumulative (
happy, 1–3); Gallup World Poll 2006–2025 waves (ladder0–10;pos_affect/neg_affect, 0–1); Integrated EVS–WVS 1981–2023 (satis_your_life, 1–10). Analysis code:analysis/closing-the-scissors/build_data.py; every number in the text is recomputed there and traced inclaims.json. - Method. Weighted OLS within each dataset. Every predictor is z-scored within country×year (year alone for GSS) using weighted z-scores, so country-level mean differences are absorbed and a coefficient answers: within a given country-year, who reports higher well-being? Bivariate = predictor alone; joint = all predictors entered together (listwise). Coefficients are semi-standardized: predictor z-scored, outcome on its native scale — not comparable across datasets.
- Weights & N. GSS
wtssps(post-stratified), N up to 70,869. Gallup World within-countrywgt, N up to ~2.95M across 164 countries. WVS within-countryweight(no population weight in this extract; figures are the average national experience), N up to 615,336 across 108 countries. Joint models use the listwise samples (GSS n=24,411; GWP n=2,131,087; WVS n=354,350). - Constructs. Evaluative outcomes: GSS
happy, GWPladder, WVSsatis_your_life. Experiential outcomes: GWPpos_affectandneg_affectonly — the GSS and WVS extracts carry no daily-affect items. The negative-affect coefficient is sign-flipped for display so higher always means better. The two constructs are never placed on a shared axis. - Missing items (shown, not faked). The Gallup extract lacks a self-rated health item and a freedom-of-choice item; both rows are blank in the Gallup panel. The WVS lacks a binary social-support item; felt financial satisfaction is used as an income-adjacent proxy. The GSS lacks both a freedom item and any experiential item. Marriage flips sign across datasets (GSS/WVS positive, GWP slightly negative) and is flagged in the text.
- Prior art. World Happiness Report (the canonical six-factor decomposition of the ladder, against which this re-runs the contest); Diener & colleagues on autonomy and well-being; the Easterlin paradox and the long debate over how far income carries subjective well-being. This coda re-derives the pattern across three independent surveys and separates the two well-being constructs throughout.