China's upside-down meritocracy: where smart graduates go to government, not the garage

On 11 July 2026, the analyst Cremieux Recueil flagged a striking statistical pattern from China: in a country where the university entrance examination remains the single most consequential rite of passage for young people, the students who score highest on it are the least likely, by several measures, to start a business. Higher Gaokao scores, the chart shows, predict lower odds of founding a company. The post, drawing on data summarised at cremieux.xyz, was blunt about the implication: a differently selected sample, with different career incentives, behaves differently. Smart people go into government.

Three days later, on 13 July, a viral incident from a residential block in China offered a near-perfect illustration of the cultural pressure that produces that sorting. A man was charged with disturbing public order after parachuting off the side of a residential building, according to the South China Morning Post. The stunt, filmed by neighbours and shared widely online, was less about daredevilry than about conspicuous individual expression in a society that has, for two generations, organised status almost entirely through credentials and institutions. The juxtaposition is the news: a society that has mastered the production of academic excellence, and a labour market that has largely forgotten how to turn that excellence into independent ventures.

The data behind the sorting

The Cremieux Recueil summary, drawn from a longer analysis on his newsletter, presents a simple but uncomfortable finding. Across cohorts and regions, Chinese graduates who sit at the top of the Gaokao distribution disproportionately take civil-service exams, enter state-owned enterprises, and pursue postgraduate study. Those lower on the same distribution are more likely to launch firms, often necessity-driven, often in tradable goods or services adjacent to existing industries. The mechanism is plausible enough that it does not require defending. The exam rewards a particular cognitive style; the rewards to that style sit inside the state.

This is not a new complaint about China. For at least a decade, Chinese-language commentary has bemoaned the "involution" of the youth labour market, the so-called neijing or "lying flat" phenomenon, and the steady drift of top graduates toward the tizheng (within-system) track. What the Cremieux data adds is a clean correlation across a wider sample. It is one thing to point at anecdotes about Peking University graduates queueing for civil-service posts. It is another to show the relationship holds in aggregate.

The structural context matters. China runs the world's largest programme of credentialed civil-service recruitment, a sprawling state-owned sector that still absorbs a meaningful share of white-collar graduates each year, and a tenure-track academic pipeline that rewards publication in domestic journals. Each of these is a respectable career in almost any society. Taken together, they constitute a different labour market from, say, the one a STEM graduate in Bangalore or Boston walks into in 2026. The Chinese state did not design this allocation on purpose. It emerged from the interaction of exam culture, the prestige ordering of employers, and an entrepreneurial environment that has cooled steadily since the late 2010s.

The parachuting man and the room for individualism

The SCMP story, with its small scale and large viral footprint, is useful precisely because it is unrelated to policy. A man jumps off a building with a parachute, bystanders film it, police charge him with disturbing public order, the clip circulates. There is no startup here, no patent, no factory. There is, instead, the spectacle of one person performing a self-authored act of risk in a public space that was designed for habitation.

In a stricter reading, the stunt is what economists would call a signal: conspicuous individual action, broadcast to a wide audience, in a society where such signals are otherwise rare for anyone outside official channels. Read alongside the Cremieux data, the two pieces illuminate each other. The exam rewards and institutionalise one kind of behaviour. When the institutions themselves have stopped absorbing graduates, the cultural pressure to perform a recognisable individual act does not disappear. It migrates.

This is not to romanticise either pillar of the story. The Cremieux summary is one analyst's reading of data that has not, to this publication's knowledge, been published in a peer-reviewed venue. The parachuting stunt resulted in a criminal charge; the social-media conversation around it has been almost entirely Western. Both data points are real and both are partial.

Why the structural read holds

A counter-explanation deserves airtime. It is possible that the apparent decline in graduate entrepreneurship is not a decline at all but a reweighting. China's industrial policy of the late 2010s and 2020s deliberately pushed capital and talent into strategic sectors – batteries, electric vehicles, solar, semiconductors, biotech, shipbuilding – through state-aligned venture and credit channels. Founders in those areas may not look like Silicon Valley founders. They look like senior engineers inside BYD, CATL, or SMIC, or like principal investigators at state-linked labs who spin out firms as adjuncts to a primary institutional affiliation. Under this reading, the top of the Gaokako distribution is not avoiding entrepreneurship. It is doing entrepreneurship inside firms that the West has not yet learned to classify as entrepreneurial.

The structural facts support both readings, which is why the question is unresolved. China continues to register more new firms per year than any other major economy, and the bulk of those firms continue to be small, low-capital, consumer-facing service businesses. At the same time, the high-value industrial achievements of the past five years have come overwhelmingly from inside, or adjacent to, large state-linked entities. The Cremieux correlation, in other words, may describe the SME economy accurately and miss most of the strategically significant firm formation. Both can be true.

What changes if the read is right

If graduate entrepreneurship is structurally suppressed by the career incentives baked into the exam and civil-service system, the implications extend beyond cultural commentary. They reach industrial policy. Beijing's stated goal of becoming a science and technology superpower depends, in part, on the assumption that the country can convert its enormous annual output of trained engineers into a comparable output of high-value firms. The Cremieux data suggests the conversion rate at the top of the distribution is weaker than the production rate. The parachuting stunt suggests, more darkly, that individuals who want to register a personal achievement may have to do it off a building.

The near-term watchpoints are concrete. China's annual civil-service recruitment figures for 2026 will show whether the drift toward the tizheng track is accelerating or stabilising. The next round of Ministry of Education graduate-employment statistics, due in the autumn, will show whether top-tier university graduates are continuing to cluster in state-owned employers. And the venture-capital deployment data from Beijing's policy banks, which is reported quarterly with a lag, will show whether capital is reaching the same graduates through corporate vehicles rather than direct investment.

The Cremieux post and the SCMP-reported parachuting charge are, on the face of it, a chart and a stunt. Read together, they suggest a society whose merit machine is running hot and whose entrepreneurship machine is running cooler than the official narrative allows. Whether the gap is a problem, a feature, or a transition state, the evidence deserves more than a glance.

Desk note: Monexus framed this piece against the standard Western reading, which treats low Chinese graduate entrepreneurship as a failure. The structural read here – that high-value firm formation may be migrating inside large state-linked firms rather than disappearing – keeps the framing evidence-led rather than moralistic.