参照系的暴政：为什么变富之后，我们并没有变得更满足
In 1974 Easterlin found that growth does not raise happiness; three decades later the same data were reanalysed and yielded the opposite conclusion. The dispute never settles because both sides argue over whether absolute income matters, while the operative variable is whom you compare yourself with. The question is not whether people compare — comparison cannot be switched off — but who sets the scale
One: Three statements that cannot all be true
In 1974 Richard Easterlin published a paper in a Festschrift for Moses Abramovitz with a question for a title: "Does Economic Growth Improve the Human Lot? Some Empirical Evidence." His answer came in three parts, and the three parts point in different directions.
First: within a single country, comparing people at one moment in time, those with higher incomes do report higher happiness. This is remarkably stable and shows up in almost every survey.
Second: comparing across countries, per-capita income and average happiness are related, but far from proportionally. Rich countries are not happier than middle-income countries by the margin their income gap would suggest.
Third: looking at one country over time. In the United States, real per-capita income rose substantially over the two and a half decades after the war, while self-reported happiness stayed roughly flat.
Put the three together and you have a contradiction. If richer people are happier (first statement holds) and the whole society has become richer (the premise of the third holds), then average happiness should have risen. It did not. The contradiction came to be called the Easterlin paradox.
The paradox deserves serious attention not because it is intellectually elegant but because it has been used repeatedly to support a very concrete policy judgment: growth is no longer worth pursuing, because its benefits are cancelled out in comparison. If that judgment holds, a great many priorities have to be reordered. So it has to survive being attacked.
Two: The data were fine; the reading was not
In 2008 Betsey Stevenson and Justin Wolfers published a long paper in Brookings Papers on Economic Activity titled, plainly, "Economic Growth and Subjective Well-Being: Reassessing the Easterlin Paradox." They did not dispute Easterlin's data. They rearranged it: income in logarithms, longer time windows, and a country sample expanded from a dozen or so into the Gallup World Poll, the World Values Survey, and Eurobarometer.
The result was clean. The logarithm of income and life satisfaction are stably and positively related both across countries and over time, with no visible saturation point. Their judgment was that the Easterlin paradox is not a phenomenon in the data but an artefact of particular samples and particular functional forms.
Two years later Angus Deaton and Daniel Kahneman did something else with Gallup data on more than 450,000 Americans: they split the dependent variable in two, separating life evaluation from emotional experience. Life evaluation rises steadily with the logarithm of income, with no turning point. Day-to-day emotional experience flattens out at a household income of roughly 75,000 dollars. This result is often compressed into "money buys happiness up to 75,000 and no further," but what it actually says is that two things are not one thing: how a person judges their life as a whole, and how that person felt yesterday afternoon, follow different mechanisms.
In 2023 Matthew Killingsworth, Kahneman, and Barbara Mellers offered a reconciliation using experience-sampling data from more than 33,000 people; the paper was titled "Income and emotional well-being: A conflict resolved." What they found: for most people — the roughly four-fifths who are already emotionally well off — well-being keeps rising with income. For the unhappy minority, there is a genuine plateau near 100,000 dollars, and beyond it well-being declines slightly.
So absolute income matters. That has to be conceded; it cannot be waved away with the cheap line that money cannot buy happiness.
Three: Neither side was measuring absolute income
The usual move here is to split the difference: Easterlin was right, Stevenson and Wolfers were right, it depends. That peacemaking conclusion conceals something more interesting.
Look again at Easterlin's first statement. He ranked people by income inside one country and compared those near the top with those near the bottom. Mathematically the variable is income; informationally it is rank. It never measured absolute purchasing power. It measured relative position. That finding was a finding about relative income, and it has spent fifty years being cited as evidence about absolute income.
Now the third statement. The time series compares a person with that person's own past. The reference point is personal history, not other people's present.
James Duesenberry named the two mechanisms separately in 1949, in Income, Saving and the Theory of Consumer Behavior. Consumption that responds to a reference group is the demonstration effect; consumption that sticks to one's own income peak is the ratchet effect. The ratchet matters because it is asymmetric: a fall in income hurts more than an equivalent rise pleases. Someone accustomed to a standard of living loses more by falling back than they gained by climbing.
So the real disagreement in this half-century argument is not whether the absolute matters. It is which reference frame dominates when. In cross-country comparisons the frame is the border; in time series it is memory; in cross-sections it is the neighbour. All three parties are right, because they are measuring along three different axes.
This also explains why the argument never ends. Change the reference frame and the same data support the opposite conclusion — and neither side will admit it changed frames.
Four: Reference frames are produced, not merely perceived
If the reference frame were only a psychological parameter, the only advice left would be "stop comparing." That advice fails, and not because people lack willpower. It fails because reference frames are supplied from outside.
Helmut Schoeck argued in Envy: A Theory of Social Behaviour (German original 1966) that the trigger for envy is comparability. No one envies an abstract billionaire; people envy the person who started near them and is now ahead. Envy requires proximity — close enough that "it could have been me." Which means the boundary of the reference group is not freely chosen by the individual. It is set by who is visible and who is comparable.
Hadas Weiss pushes the mechanism up to the level of identity in We Have Never Been Middle Class. Her argument is that "middle class" was never a statistical position but a project: human-capital investment, a narrative of mobility, a self-enterprise financed by debt. Its most important function is not to describe where you are but to keep supplying a position just above your reach as the point of reference. Every failure to arrive is then read as personal mismanagement rather than structural outcome. Here the reference frame is not a bias. It is a governing device.
Thomas Piketty supplies pressure from the other direction in Capital in the Twenty-First Century: when the return on capital persistently exceeds the growth rate, the top moves faster than wages do. Align your position with the top and the gap widens even as your absolute income rises. The reference point outruns you. That is arithmetic, not attitude.
The following is an inference, not an empirical finding. These mechanisms were once bounded by the radius of visibility — you can only envy what you have seen. The contemporary information environment has stretched that radius to the global top, so the reference group has gone from dozens of people to tens of millions. If Schoeck is right about comparability, then widening the radius of visibility should systematically raise the reference point without changing anyone's absolute income. This has not been cleanly tested, so I hold it as a direction to investigate rather than a conclusion.
Five: What can be changed is not comparison but its dimensions
So the question should not be whether people ought to compare. Comparison is a by-product of the capacity to rank; it cannot be switched off. The question is whether there is only one scale, and who sets it.
Three tests:
- Are the dimensions selectable? If a person is evaluated on a single axis — income, title, follower count — comparison is necessarily zero-sum: every step you rise is a step someone else falls. Multiple dimensions do not abolish comparison, but they make rankings incommensurable, which blunts the damage of a single ladder.
- Is the gap attributable? If a gap traces back to specific choices and effort, comparison produces motivation. If it traces back to nothing anyone could have done — birthplace, industry cycle, asset prices — comparison produces only shame. Shame does not drive action; it drives avoidance.
- Can the reference group be exited? An exitable reference frame is a choice; an inexitible one is a structure. A person can unsubscribe from an account but cannot leave the pay structure of their industry.
When all three hold, comparison is information. When none hold, comparison is an instrument of punishment. And the trouble with most modern situations is that the third is increasingly unsatisfiable, the first is narrowing, and the second is growing murkier.
The limits need stating too. First, under absolute poverty, relative comparison is a luxury proposition. What Poor Economics documents repeatedly is that below a certain line the problem is not rank but certainty: the next meal, the next harvest, the next illness. Telling those households to stop comparing translates a structural problem back into a problem of attitude. Second, the Deaton-Kahneman curve at 75,000 dollars and the unhappy minority identified by Killingsworth and colleagues both indicate that absolute security cannot be replaced by a discussion of relativity: reference frames can explain why getting richer does not satisfy, but they cannot explain why poverty genuinely hurts. Third, anyone who announces that they have stopped comparing has usually just swapped in a more concealed reference frame — against an ideal self, or against people who do not compare. The frame does not disappear; it relocates.
Morgan Housel's The Psychology of Money contains a line to the effect that the great irony of wealth is that it only ever shows you the part you do not yet have. That line holds not because human nature is greedy but because the reference point has been institutionally arranged to move faster than income. What needs changing is not the impulse to compare but the single scale that has been erected and left standing alone.
Six: Back to the three statements
Easterlin, Stevenson and Wolfers, Deaton and Kahneman: none of them was wrong. Each was measuring along one axis and mistook that axis for the only one.
What the argument actually exposes is not the old question of whether money matters but a methodological fact: almost every measurement of well-being quietly selects a reference frame and then leaves that selection out of the conclusion. Hence the same data can support two policies, and both can claim empirical warrant.
If there is an operational takeaway, it is this. Before arguing about whether something is enough, establish what it is enough relative to. Without that question, every number that follows is empty.
