Fair Play Study
@zerisinyu
A data story · Chinese General Social Survey 2023
China removed every limit on having children, then started paying people to have them. Births kept falling anyway. We asked the data why — and 4,889 young adults pointed somewhere unexpected.
“When will you have a baby?”
is the question young Chinese adults get at every family dinner. This story is about a different one, asked quietly by a national survey:
“Do you believe your child
would get a fair chance in this society?”
Local governments are trying, sincerely, to pay for babies. One mid-sized city — call it City M — offers a package worth roughly $13,600 for a second child and $23,300 for a third, counting subsidies, housing credits and leave.
A woman we’ll call Su Fang recently had her third child there. State media praised her as a role model. Behind the headline: the pregnancy was an accident, her husband wanted a son after two daughters, and she nearly tested the baby’s sex. Her verdict on the policy that celebrated her —
Press clipping · 新闻剪报City M · 2024
“Give me five million yuan and I still won’t have another one.”
— “Su Fang”, mother of three, as told to state media
She is not an outlier. Here is what happened to births while the incentives scaled up:
The more was allowed, the fewer arrived
Annual births in China, millions
Source: National Bureau of Statistics of China. Dashed markers: universal two-child policy (2016) and the removal of all birth limits (2021).
Nearly half as many babies in eight years, through the loosest birth policy in four decades. If it were mainly about permission, this line would bend up. If it were mainly about money, the subsidies should at least have slowed it down.
So we looked somewhere else: not at what young people have, but at what they believe.
You first ↓
Forget policy, money, timing. In your ideal life, how many children would you have?
“How many children would you ideally have?”
Share of respondents by answer · CGSS 2023, born 1980+
N = 3,161 Millennials · 1,661 Gen Z with a valid answer.
Millennials (born 1980–94) still mostly picture a family with two children. Their average ideal: 1.79.
Then Gen Z answers the same question — and the whole curve slides left. Average ideal: 1.22. Between two generations raised a decade apart, half a child disappears from the imagined future.
Look at the leftmost bar. 27.4% of Gen Z say zero. Not “one, later, maybe” — none. Among Millennials it was 8.9%. This isn’t a delay. It’s a different answer.
“Zero” is not spread evenly. It concentrates exactly where the costs of childbearing — and the doubts about what a child’s life would look like — concentrate.
Share who want no children at all
Unadjusted shares of valid answers within each group. Groups overlap (an urban Gen-Z woman appears in three rows).
Women say zero at twice the rate of men. City dwellers at nearly twice the rate of rural residents. These are the people for whom a child most visibly collides with a career, an apartment, a life already stretched thin — which sounds, again, like money.
Except the data keeps refusing that easy story.
Here is the strange part. Ask these same young adults about the economy’s outcomes, and they are surprisingly at peace:
The dream isn’t dead. So we split “fairness” into two different questions — because they turn out to have very different answers.
💰 Distributive fairness
“Is the gap between rich and poor acceptable?”
About outcomes — who has what, today
🪜 Procedural fairness
“Do children get an equal shot? Does effort decide success?”
About the rules — whether the game is worth entering
What moves the wish for children — and what doesn’t
Change in ideal number of children per one-point increase in each belief, holding age, income, education, marriage, residence and existing children constant
OLS on 20 multiply-imputed datasets, robust SEs, N = 4,889. Bars are 95% confidence intervals.
Each dot is one belief’s relationship with wanting children, after holding constant the usual suspects — income, education, marriage, age, where you live. Right of the line: believing this goes with wanting more children.
“Society is fair.” Each step up this belief goes with wanting 0.10 more children — modest but unmistakable, the most reliable signal in the model.
“Effort decides success.” Same direction, similar size. People who still believe the ladder works picture larger families.
And “the income gap is fair”? Nothing. The dot sits on the zero line. How people feel about who has what tells us almost nothing about whether they want children.
That is the finding in one picture: Fair play matters. Fair pay doesn’t. Young Chinese seem to decide about children less by the wealth around them than by whether the rules feel worth trusting a child to.
Average out 4,889 people and you flatten the most human part of the story: which fairness a person needs depends on where they stand.
Urban and rural China read fairness differently
Same model, estimated separately within each group
Filled dots: p < 0.05. Hollow dots: not statistically distinguishable from zero.
Urban residents respond to the general climate — “is this society fair?” (+0.12). Living amid neijuan 🌀, the grinding competition where effort inflates but rewards don’t, what they need is diffuse: faith in the system itself.
Rural residents respond to something more specific: meritocracy (+0.20, their strongest signal). For a village family, one belief carries everything — that the gaokao 📝 can still lift a child out. Where that ladder feels real, children feel worth raising.
And a quiet paradox: rural respondents believe more in meritocracy than urban ones (1.78 vs 1.67 on our index) — and want more children (1.79 vs 1.52). Hope and fertility travel together.
Split by gender instead, and the signal weakens for women — fairness beliefs move men’s intentions almost twice as strongly. A fair society may be a sufficient signal for a man; a woman still faces the motherhood penalty and the second shift inside that fair society.
For women, societal fairness looks necessary but not sufficient. That asymmetry is one of the most policy-relevant patterns in this data — and one of the least discussed.
Now you again ↓
Overall, how fair does today’s society feel to you?
The gradient, in its rawest form
Average ideal number of children at each answer to “is society fair?”
Unadjusted means; dot size = number of respondents. This is a sketch — the regression above is the evidence.
Smaller patterns, offered as-is — unadjusted averages, not causal claims. Each one is a thread someone should pull.
💍 Marriage
1.83 vs 1.29
Ideal children, married vs unmarried respondents — the marriage cliff may matter as much as the baby cliff
🪪 Hukou
1.71 vs 1.41
Rural-hukou vs urban-hukou holders, wherever they live now — the household registration follows you
🎓 Education
1.90 vs 1.45
Respondents without vs with college education — schooling and family size trade off here too
This is one survey, one country, one year. It measures intentions, not births. And it shows associations, not causes: we cannot prove that restoring faith in fairness would raise fertility — only that the two travel together, robustly, in ways money beliefs do not.
But if the association runs even partly in the direction it points, it suggests something subsidies cannot buy. Pronatalist policy has treated children as a purchase to discount. Young Chinese seem to treat them as a bet on the future — and you don’t change a bet by changing the price. You change it by changing the odds.
Questions we’re left holding — pick the one that stays with you:
Chinese General Social Survey (CGSS) 2023, restricted to respondents born 1980 or later (Millennials + Gen Z), N = 4,889. CGSS is a nationally administered face-to-face and telephone survey; microdata requires free registration and is not redistributed with this project. Annual births are from the National Bureau of Statistics.
Outcome: “ideal number of children absent any policy restriction” (0–10). General fairness: 5-point scale. Meritocracy index: mean of two binary items (“hard work is rewarded”, “effort, not family background, decides success”), recoded so higher = stronger belief. Income fairness: binary, higher = current income differences are fair.
CGSS 2023 uses a split questionnaire: each fairness module reached only part of the sample (53–73% missing by design). We used multiple imputation — 20 datasets via iterative Bayesian ridge imputation with posterior sampling — and pooled every estimate with Rubin’s rules. Diagnostics support missing-completely-at-random by design.
Main estimates: OLS with HC3 robust standard errors, controlling for age, gender, education, marriage, log income, urban residence, hukou status and existing children. Robustness: Poisson and negative binomial count models agree in sign and significance. Subgroup contrasts come from stratified models; formal interaction tests do not survive false-discovery-rate correction, so stratified patterns are descriptive, not confirmed differences.
Cross-sectional data → associations, not causation. Stated intentions ≠ actual births. Effects are modest (β ≈ 0.10 children per scale point). Reverse causality and omitted variables (e.g. general optimism) cannot be ruled out. The two reader polls on this page store nothing and send nothing — they exist only in your browser, for you.
Every number and chart on this page is generated by a pipeline from raw microdata — code, paper and instructions on GitHub. Charts: D3.js. Scroll: Scrollama. Palette checked for color-vision-deficiency separation and contrast.