What this research found
The Easterlin paradox claims that richer people are happier at any moment but that a country growing richer does not become happier over time. Testing that directly meant estimating both slopes from the same data with the same functional form: the between-country gradient in the 2019 cross-section of 148 countries, and the within-country gradient from a country fixed-effects panel of 160 countries over 2011–2022. The within-country slope came out at 1.139 points per log-unit of income against a between-country slope of 0.750 — about 52% steeper, not attenuated — so the paradox is not supported over this window.
- In the 2019 cross-section of 148 countries, life satisfaction rises sharply with income: Pearson correlation 0.787 with log GDP per capita, Spearman 0.799, and a bivariate slope of 0.750 ladder points per log-unit (SE 0.049, p ≈ 2.1 × 10⁻³², R² = 0.619). A doubling of income corresponds to about a 0.52-point gain on the 0–10 ladder.
- In the 2011–2022 panel of 1,626 country-year observations across 160 countries, country fixed effects give a within-country slope of 1.139 (cluster-robust SE 0.235, p ≈ 1.3 × 10⁻⁶). Adding year fixed effects to absorb common global shocks raises it to 1.343.
- The within slope exceeds the between slope by 0.390 points, a ratio of 1.52. The formal comparison gives z = 1.62 with a two-sided p of 0.105, and the one-sided test in the Easterlin direction returns p = 0.95 — no support for attenuation in either specification.
- Dropping the pandemic years lowers the within slope to 0.911 on 2011–2019, but it still sits above the between-country 0.750 and is not significantly different in either direction (z = 0.67, two-sided p = 0.51), so the conclusion does not hinge on the COVID-19 rebound.
- The result is not driven by a few fast-growing giants: excluding China and India raises the within slope to 1.208, and dropping countries observed in only one year leaves it unchanged at 1.139.
- Income enters well-being logarithmically rather than linearly. The log specification fits the 2019 cross-section better than the level specification (R² 0.619 against 0.560), matching the concave shape reported in the existing literature.
How it was done
Two public panels were merged on ISO3 country code and calendar year: national mean Cantril life-ladder scores compiled by Our World in Data from the Gallup World Poll, and World Bank GDP per capita at purchasing-power parity in constant 2021 international dollars. Thirteen Our World in Data aggregates and forty-four World Bank regional and income aggregates were stripped out first, Kosovo's non-standard code was remapped so it matched across sources, and the audit found no duplicate country-year cells; 164 countries matched, with Taiwan present only in the happiness data and 52 small states only in the income data. The between-country gradient was estimated by ordinary least squares on 2019, chosen as the most recent pre-pandemic year with broad coverage. The within-country gradient came from least-squares dummy-variable estimation with country and then country-plus-year fixed effects, standard errors clustered by country, and the two slopes were compared with a z-test on their difference alongside pre-pandemic and leave-out robustness checks.
Data sources
- Our World in Data — self-reported life satisfaction (Cantril ladder) from the Gallup World Poll and World Happiness Report, 2,270 country-year rows across 178 entities
- World Bank World Development Indicators — GDP per capita, PPP, constant 2021 international dollars (series NY.GDP.PCAP.PP.KD), 1990–2023
- Stevenson & Wolfers, Brookings Papers on Economic Activity 2008(1):1–87 — the principal challenge to the paradox
- Easterlin et al., PNAS 107:22463–22468 (2010) — the reaffirmation of the paradox on a 37-country panel
Limitations
Twelve years cannot adjudicate a claim about multi-decade trends, so the estimate captures medium-run dynamics rather than the long-run steady state Easterlin argued about; the authors read the result as evidence that income and life satisfaction move together at a business-cycle-to-decade horizon, not as a refutation of the long-run claim. The ladder is also an ordinal, self-anchored scale treated as cardinal in the regressions, and the retrieved income series uses the 2021 rather than the specified 2017 purchasing-power-parity base year.
How this research was produced
K-Dense Web planned and ran this economics investigation end to end — gathering the sources, carrying out the analysis, producing the figures, and drafting the report. The full session transcript, including every intermediate step, is available to view.
