Chapter 3 — Measurement: GDP vs well-being¶
This chapter examines what GDP measures, what it leaves out, and why alternative well-being frameworks capture dimensions that standard economic metrics ignore. By the end, the reader will understand the difference between economic activity and real well-being, know the Easterlin paradox and the Kahneman–Killingsworth debate on the relationship between income and life satisfaction, and have a framework for evaluating AI's impact on well-being beyond what GDP shows.
Prerequisites
This chapter assumes you know the concepts introduced in Chapter 2 — AI as an electrical technology.
When we want to know whether a country is "better off" than before, or whether a technology is improving people's lives, we need to measure it. The indicator used most often for that purpose is GDP. The problem is that GDP measures what it measures well — economic activity — but that is not the same thing as well-being.
1. What GDP measures and what it leaves out¶
GDP sums the value of all goods and services produced in an economy over a period of time. As a metric of economic activity, it is precise and comparable across countries. The problem is not what it measures, but what it does not measure.
What GDP leaves out¶
Unpaid work. Caring for older people, raising children, maintaining a household and community volunteer work are not included in GDP because they do not pass through the market. OECD and World Bank estimates place the value of this unpaid work at between 30 and 40% of official GDP in developed economies OECD (2020), making it the largest category of production that the indicator ignores entirely.
Distribution. GDP per capita is an average that can hide extreme inequalities. A country where 1% of the population captures 90% of growth can have rising GDP per capita while most of the population experiences no improvement at all. The arithmetic mean says nothing about how what it measures is distributed.
Quality of services. An education system that produces more graduates with lower skills can report higher GDP — more measured economic activity — while producing a worse outcome in terms of real human capital. The quantity of activity and the quality of the outcome are different things.
Negative externalities. Pollution, depletion of natural resources and environmental degradation do not appear in GDP as costs. A country can grow while destroying its natural environment without subtracting anything from the indicator. When that destruction requires spending on remedies — hospitals, water purification, infrastructure repair — that spending adds to GDP instead of subtracting from it.
Security and social cohesion. The absence of violence, institutional trust, social stability and the ability to plan for the future have enormous value to people and have no direct representation in GDP.
2. What well-being measures that GDP does not capture¶
The well-being frameworks developed by economists, psychologists and international organizations over recent decades converge on similar dimensions.
Goods and services that pass through markets.
Transactions comparable across countries and periods.
Whether the economy accelerates, slows or enters recession.
Who improves and who is left out of the average.
Health, time, security, social support and environment.
Whether people feel their lives are actually improving.
The average rises, not the majority.
Growth can worsen air quality, time or health.
More output does not guarantee purpose or security.
Material dimensions¶
Income and wealth, employment and job quality, housing, accessible health services and quality education: these are the dimensions where GDP has the greatest predictive capacity, although even here distribution matters more than the average.
Non-material dimensions¶
Work–life balance, social relationships and community support, civic participation and governance, personal security, environmental quality and subjective well-being: these are the dimensions that diverge most from GDP and where the paradox of "growing without improving" appears most often.
3. Subjective well-being and when it diverges from material well-being¶
Subjective well-being — what people report when asked how satisfied they are with their lives — is a complicated but relevant metric. Its limitations are well known: people adapt to circumstances (what once seemed intolerable becomes normal), comparison anchors matter enormously (satisfaction depends heavily on whom you compare yourself with), and surveys have cultural and momentary biases. Recent research also distinguishes hedonic well-being (absence of distress, immediate pleasure) from eudaimonic well-being (sense of purpose, autonomy, fulfilment), two dimensions that can move in different directions under the same material changes OECD (2013).
Despite those limitations, subjective well-being data capture something GDP cannot: whether people feel their lives are good, regardless of what aggregate indicators say. Data from the World Happiness Report 2024 illustrate the divergence: Finland, with a life-satisfaction score of 7.7 out of 10, topped the ranking for the seventh consecutive year despite not being the European country with the highest GDP per capita, while Spain, with an income level comparable to several Nordic countries, sits around 6.5. The difference is not explained by income but by variables such as institutional trust, social support networks and perceived freedom to make life decisions.
The Easterlin paradox¶
In the 1970s, economist Richard Easterlin documented a pattern that remains relevant Easterlin (1974): within a country, at a given point in time, people with higher incomes report higher subjective well-being.
But across countries over time, growth in GDP per capita does not produce the same kind of increase in subjective well-being once a basic threshold has been passed. The effect of relative position — how much you have compared with those around you — can dominate the effect of absolute income. Easterlin's research, which he continued refining until his death in 2024, generated decades of empirical debate over whether the threshold exists and where it lies. In 2023, that debate produced a methodologically important result: Daniel Kahneman and Matthew Killingsworth published an adversarial collaboration designed to reconcile their previously contradictory conclusions, with Kahneman arguing that hedonic well-being stopped rising with income above $75,000 per year and Killingsworth finding continued growth with no ceiling.
The joint analysis found that both were right for different subpopulations: for most people, well-being measured in real time continues to rise with income with no apparent ceiling, but for approximately the 20% who already report low well-being, income above that threshold produces no appreciable additional improvement. The Easterlin paradox and the ceiling hypothesis are compatible because they apply to different people.
This has direct implications for how we evaluate AI's impact: if AI increases GDP but concentrates the gains among a few, the effect on subjective well-being can be neutral or negative for the majority even while the aggregate indicator improves.
When material and subjective well-being diverge most¶
Comparative studies point to the situations and contexts where the divergence is greatest:
Rising inequality. When economic growth is not shared evenly, the subjective well-being of those left behind can deteriorate even while the average rises.
Job insecurity. Even when income remains unchanged, perceived employment instability consistently reduces subjective well-being. Fear of losing what one has weighs more heavily than the value of having it.
Loss of meaning. Work provides not only income but also structure, social identity and a sense of purpose. When automation removes tasks without replacing them with others that have the same subjective value, the effect on well-being can exceed what income statistics show.
Commuting time and living conditions. Time lost to long commutes, noise pollution, excessive urban density and lack of access to nature consistently erode subjective well-being, although they rarely appear as priorities in standard economic metrics.
For the 20% of people with chronically low well-being, income above ~$75,000 produces no additional improvement in daily emotional state. The plateau exists for this group.
For the satisfied majority, real-time well-being continues to rise with income with no apparent ceiling. The plateau does not apply to them.
If AI concentrates gains among those who already have an income and well-being advantage, the impact on the majority's subjective well-being can be neutral or negative even if aggregate GDP improves.
4. Alternative frameworks that try to capture more¶
Several efforts have tried to go beyond GDP as an indicator of progress.
Human Development Index (HDI) (UNDP): combines life expectancy, education and gross national income per capita. More complete than GDP alone, but it still does not capture distribution or subjective well-being.
Genuine Progress Indicator (GPI) (Talberth et al., 2007): starts from personal consumption, adds the value of volunteer and household work, and subtracts the costs of inequality, crime, pollution and resource depletion. It produces very different results from GDP in developed countries: states with high GDP per capita can rank very low in GPI once their negative externalities are deducted, while states with lower economic activity but less inequality and environmental degradation move up the ranking. The divergence between the two indicators has widened since the 1970s in most advanced economies Talberth et al. (2007).
OECD Better Life Index (OECD, 2020): eleven dimensions that each user can weight according to their priorities. It explicitly recognizes that there is no single correct definition of well-being.
Gross National Happiness: Bhutan's framework, which incorporates psychological well-being, time use, community vitality and cultural resilience. Radical in its breadth, difficult to compare internationally.
None of these frameworks has displaced GDP as the dominant indicator because GDP has real advantages: it is precise, comparable, updated frequently and politically neutral in its methodology. But alternative frameworks show what is lost when GDP is used as a proxy for general well-being.
The next chapter applies this distinction to the specific case of AI: where its impact on productivity is already appearing, why GDP sees it late, and which early signals are more indicative of what is happening.
Frequently asked questions¶
What are the main dimensions of well-being that GDP does not measure? GDP measures market economic activity. What it leaves out includes unpaid work (caregiving, child-rearing, volunteering, which the OECD estimates at 30–40% of official GDP), the distribution of that activity across the population (GDP per capita is an average that can hide extreme inequality), the quality of services (more university graduates does not imply more human capital), negative externalities such as pollution, and subjective well-being variables such as security, institutional trust or sense of purpose.
What does the Easterlin paradox say about income and happiness? Easterlin documented that within a country and at a given point in time, people with higher incomes report higher subjective well-being. But across countries over time, growth in GDP per capita does not produce the same increase in subjective well-being once a basic threshold has been passed. The paradox proposes that the effect of relative position (how much you have compared with your neighbors) can dominate the effect of absolute income. Later research, including an adversarial collaboration between Kahneman and Killingsworth published in 2023, suggests that both patterns can be true for different subpopulations.
How does HDI differ from GDP, and why is it still not enough? The Human Development Index combines life expectancy, years of education and gross national income per capita. It is more complete than GDP alone because it includes two non-economic dimensions. But it still does not capture the distribution of those indicators within a country, subjective well-being, environmental quality or variables such as personal security and social cohesion. The inequality-adjusted HDI (IHDI) tries to address the first problem, but the other dimensions remain outside it because of measurement and international-comparability difficulties.
Why has no alternative indicator displaced GDP as the dominant measure? Because GDP has real advantages that alternative frameworks do not offer: it is precise in what it measures, can be updated frequently, is internationally comparable with a standardized methodology and is politically neutral in its methodology (it takes no position on which dimensions are most important). Frameworks such as GPI, HDI or the OECD Better Life Index require decisions about what to weight more heavily, making them conceptually richer but harder to use as reference points in comparative economic policy.
How does AI automation affect well-being beyond income? If AI automates tasks without creating others with equivalent subjective value, the effect on well-being can be negative even if income is maintained. Work is not only a source of income but also of social identity, temporal structure and sense of purpose. Subjective well-being studies show that job insecurity consistently reduces well-being even when income remains unchanged, because fear of losing what one has weighs more heavily than the current value of having it. If AI displaces work without that transition being managed, the impact on subjective well-being can exceed what income data indicate.
5. References¶
Core sources
| Key | Source | Brief description |
|---|---|---|
| R1 | UNDP — Human Development Reports (UNDP) | Conceptual framework and data for the Human Development Index. Annual publication with life expectancy, education and income data for more than 190 countries. |
| R2 | OECD (2020) — How's Life? Measuring Well-being (OECD) | Better Life Index framework and comparative analysis of well-being across 11 dimensions. Includes estimates of the value of unpaid work in developed economies. |
| R3 | Easterlin, R.A. (1974) — Does Economic Growth Improve the Human Lot? Some Empirical Evidence (in Paul A. David and Melvin W. Reder, eds., Nations and Households in Economic Growth) | Original formulation of the paradox between GDP growth and subjective well-being. Starting point for decades of empirical debate over the basic-needs satisfaction threshold. |
| R4 | Stiglitz, J., Sen, A., Fitoussi, J.-P. (2009) — Report by the Commission on the Measurement of Economic Performance and Social Progress (Eurostat) | Report of the commission convened by Sarkozy. Systematizes GDP's limitations and proposals to incorporate distribution, sustainability and subjective well-being. |
| R5 | Penn World Table (University of Groningen) | Comparative database of GDP, productivity and material well-being adjusted for purchasing power parity for more than 180 countries since 1950. |
| R6 | Talberth, J., Cobb, C., Slattery, N. (2007) — The Genuine Progress Indicator 2006 (Redefining Progress) | GPI methodology and data for the United States. Documents the widening divergence between GDP and GPI since the 1970s in advanced economies. |
| R7 | OECD (2013) — OECD Guidelines on Measuring Subjective Well-being (OECD) | Methodological framework for measuring subjective well-being, including the distinction between hedonic (affective evaluation) and eudaimonic (functioning and purpose) dimensions. |