The Village Factories the State Did Not Plan

Published October 4, 2026

Yasheng Huang, an economist at M.I.T., has spent two decades irritating the standard story of Chinese growth. The story this book's earlier chapters share with both admirers and skeptics is that growth came from the state's ability to move resources, direct investment, and coordinate a continental economy. The Party built the infrastructure, managed the currency, sometimes disciplined the banks, opened the special economic zones, and set the conditions for four decades of expansion. The argument then becomes whether Beijing did that management well or badly. Huang says the premise is wrong.

His research, laid out across several books and most recently in The Rise and Fall of the EAST, traces the most explosive period, the 1980s, to rural entrepreneurship that flourished because the state was weak and disorganized in the countryside. Township and village enterprises, the hybrid outfits that drove rural industry in the reform years, were not products of a plan. They were products of the state's absence. Rural entrepreneurs used the gap between what Beijing wanted to control and what it could actually reach. They built factories in villages where no inspector came, hired workers off the books, sold into markets the planning system did not watch, and produced growth the state later claimed and had not directed.

The awkward piece for the poverty campaign is specific. The Targeted Poverty Alleviation drive of 2013 to 2020, three million cadres, careers tied to named households, is evidence of institutional competence in an earlier chapter. Huang's research says the problem that campaign solved was partly a problem the state made in the 1990s, when Beijing shifted from rural decentralization to urban-focused state capitalism. Investment piled into state-owned enterprises, coastal cities, and heavy industry, and starved the rural economy that had been generating the fastest growth. The 100 million people in extreme poverty in 2013 were not, in his account, a leftover that growth had simply failed to reach. They were casualties of a choice that put state-directed city development ahead of the rural market that had been working. If he is right, the competence of the later campaign has to be weighed against the choices that made the campaign necessary. The same system that sent three million cadres to fix rural poverty is the system that let rural poverty persist while GDP quadrupled. The chapter does not throw out the earlier win. It says the win was partly a correction of a self-inflicted wound.

Desmond Shum's Red Roulette, published in 2021, describes a layer no academic chart fully holds. Shum was a businessman in Beijing's political-commercial elite from the early 2000s through the 2010s, connected through his then-wife Whitney Duan to networks that included senior party officials and their families. The memoir is not an analysis. It is an account of how deals actually get made. The rules, as he describes them, are not the rules in party documents or regulatory frameworks. They are relational, unwritten, and they change without notice. An introduction to the right official's family member opens a development project. A shift in faction closes it. The same behavior that earns rewards under one patron earns a prison sentence under the next. Whitney Duan disappeared in 2017. Security services took her. Shum learned about it secondhand. It has never been publicly explained.

What that account complicates is the assumption that the system runs through institutions: incentives, information flows, the gap between a principal and an agent. Shum's testimony says that at the highest levels, where major deals happen, the machinery is personal. Patrons and clients run on trust, reciprocity, and mutual vulnerability, not on an organizational chart. The tech crackdown can still be read as overlapping regulators and fused political and economic authority. Shum's world adds a layer: targets selected by the factional ties of the people who controlled the companies. Jack Ma's vulnerability, on that reading, was not only that Ant Group posed a systemic financial risk. It was that his network included people on the wrong side of a factional consolidation. The two readings predict the same result in that case. They would not always.

Every chapter in the book has used Chinese economic numbers, and every chapter has noted that the numbers are unreliable. John Balding's research on the consumer price index documented a systematic suppression of reported inflation by about one percentage point a year between 2000 and 2011. Over a decade, understating inflation by that margin cuts the real value of reported GDP growth by 8 to 12 percent. The growth miracle in the official statistics overstated real output gains by roughly a tenth. The China in the data grew faster than the China in the physical world. The New York Fed noted that the volatility of China's reported GDP growth was unbelievably low. Real economies swing. Business cycles produce strong quarters and weak ones. China's reported GDP moved with a smoothness no market economy and no planned economy has achieved. The signature was not of an economy being measured. It was of a number being managed.

The China Beige Book, which surveys businesses without using government data, has consistently reported a more pessimistic picture. Employment, consumption, and investment diverge from official figures in one direction: things are worse than Beijing says, by a margin that changes by sector and region and does not go away. Provincial fraud compounds it from below. Inner Mongolia admitted in 2017 to inflating industrial output by 40 percent. Liaoning acknowledged years of fabricated data. Tianjin revised its GDP down by a third. When the inputs are made up at the province, the national total inherits the distortion. The fraud was not random noise. It was systematically upward, because the cadre system rewarded officials for hitting growth targets, the same system praised for poverty work when the target could be checked. Satellite pictures checked the poverty homework. Nobody checked the GDP homework for years. The mechanism is the same. When the target is measurable and verifiable, the system delivers. When the target rewards the reported number rather than the verified outcome, the system fabricates. The difference is whether anyone is checking.

The Double Reduction policy, examined in the book's research, destroyed a 100 billion dollar private tutoring industry overnight. The demographic rationale Beijing offered rested on fertility and education-cost data subject to the same distortions. Officials decide on numbers that flatter the last round of decisions. The errors compound until reality forces a correction the data should have triggered years earlier. The chapter's conclusion is not that earlier verdicts flip. It is that the confidence intervals should be wider, because the economic base under trade, industry, the Belt and Road, and demography is systematically more optimistic than reality.

One of the cleanest Western stories is that censorship kills innovation. Creative work needs free expression, free expression needs the absence of censors, therefore a censored society cannot make world-class creative work. Recent products refuse the prediction. Black Myth: Wukong, from the studio Game Science, sold 20 million copies within months of its August 2024 release. It was among the most-played games on Steam at launch and won multiple international awards. It drew on Journey to the West, a classical novel every Chinese child knows, and competed with the best Western studios. It was made under censorship. The studio worked inside China's content rules. Ne Zha 2, an animated film released in January 2025, earned 1.69 billion dollars at the Chinese box office alone, among the highest-grossing films ever in a single market. It drew on Chinese mythology, was made by a Chinese studio, and was marketed mainly to Chinese audiences. A domestic market of 1.4 billion people can sustain blockbuster economics with no international release at all. Genshin Impact, from the Shanghai studio miHoYo, generated billions in global revenue as a live-service game, one that keeps earning after purchase through ongoing play, and drew tens of millions of players across Asia, Europe, and the Americas. Its look and its systems competed with Japanese and American work in a genre those industries had dominated.

The censorship is real. Game Science reportedly navigated content restrictions through development. Studios censor themselves to avoid regulatory risk. Whole categories stay off limits: politically sensitive history, certain depictions of violence, anything that challenges party narratives. The clean prediction was that these products could not exist. They exist. A system can suppress political speech and still, through the size of the home market, enable commercially successful creative work. TikTok is the same paradox from another side. It is one of China's most successful commercial exports, a demonstrated intelligence vulnerability with potential access to behavioral data on more than a billion foreign users, and a recommendation engine that shapes what hundreds of millions of people see. The censorship that constrains Chinese creators did not stop ByteDance from building one of the most influential media platforms on earth. The political control that many observers expected would make Chinese tech uncompetitive produced a product the United States spent years trying to ban. The domains where censorship binds, political expression, historical narrative, criticism of the regime, may not overlap completely with commercial technology, consumer products, and applied engineering.

Jessica Chen Weiss's research complicates the picture of foreign policy as a decision at the top that the diplomatic corps then carries out. Popular nationalist mobilization, through online platforms, street protests, and consumer boycotts, increasingly constrains leaders rather than simply empowering them. The usual assumption is that Beijing turns the volume up when it needs domestic support and down when diplomacy needs room. Weiss showed it is not that clean. Once the sentiment is mobilized, it generates its own momentum. A concession that looks like a violation of nationalist expectations triggers a backlash the regime has to manage. Allowing or repressing a protest is a signal, and the signal cannot be perfectly set. A protest that serves a purpose on Monday can be an embarrassment on Tuesday. Online nationalism, on platforms the state controls but cannot fully direct, outruns the bureaucracy.

Hu Xijin, former editor-in-chief of the Global Times, spent years amplifying hardline positions that matched Beijing's tougher foreign policy. More recently he has warned that unrestrained nationalist pressure could damage diplomatic interests. When a professional nationalist says it has gone too far, the job has shifted from switching a tool on and off to managing a wildfire. The 2012 anti-Japanese protests are the starkest case. Demonstrations first encouraged by state media, to signal displeasure over the Senkaku islands, which China calls the Diaoyu, escalated into riots that destroyed Japanese-brand car dealerships, factories, and restaurants across dozens of cities. The economic damage fell on Chinese workers at Japanese firms and on Chinese owners of Japanese cars. The protests achieved no diplomatic objective. They embarrassed Beijing and forced a crackdown on the sentiment the state had cultivated. Later boycotts of South Korean, Australian, and Philippine businesses repeated the pattern at smaller scale, with less diplomatic return and higher domestic cost. Influence over information is not control. The sentiment, once shaped, acts back on the shaper. When Beijing misjudges a foreign response, the miss may not only be a failure to model how democracies react. It may also be a domestic audience whose expectations the leaders cultivated and can no longer fully manage.

Eric Li, a venture capitalist and political theorist, makes an argument most Western analysts will not take up on its own terms. His claim is not that the system is imperfect but improving. It is that one-party governance is structurally better than multiparty democracy for a country of China's scale, and that the evidence supports the claim. The first piece is self-correction. The Party, he says, self-corrects in dramatic fashions. The shift from Mao's Cultural Revolution to Deng's market reforms is his paradigmatic case: a 180-degree reversal without a change of regime, without a revolution, without a civil war. Democracies correct through elections, which are slow, noisy, and tied to short cycles. The second piece is selection. He describes the Organization Department, the Party's internal personnel office, as a giant human resource engine that evaluates, promotes, and moves officials on performance metrics from the village to the Standing Committee. Elections, in his contrast, select on media skill, fundraising, and ideological appeal rather than on demonstrated administration. The chapter says this argument is internally coherent and fails against specific cases. The Great Leap Forward, the Cultural Revolution, and the zero-COVID lockdowns are places where self-correction did not turn on, or turned on only after catastrophic damage. The Big Fund corruption is a place where the selection apparatus promoted officials who stole billions. The mechanism requires, at some point, that someone at the top recognize the error. When the person at the top is the error, the mechanism breaks. The failures make the argument incomplete, not incoherent. When the correction does work, as in the poverty campaign and the semiconductor pivot after the corruption cases, the results are formidable. No theory, including Li's, predicts in advance which cases will trigger a correction and which will trigger doubling down.

Branko Milanovic's inequality data reframes the poverty numbers in a global setting. China's alleviation was the most dramatic piece of a broader convergence. Extreme poverty fell across Asia, parts of Africa, and Latin America in the same decades. Globalization, technology transfer, urbanization, and export-led growth ran through countries with radically different politics. Vietnam reduced poverty under one-party rule. India reduced it under democratic government. Bangladesh reduced it under a hybrid that does not sort cleanly. The common piece was not regime type. It was joining global trade networks during a period of expanding demand for labor-intensive manufactured goods. China's scale was exceptional. No other country moved as many people above the line as fast. Exceptional scale does not automatically prove an exceptional mechanism. If the main driver was sustained GDP growth from global market integration, the Party's contribution was creating conditions for growth, infrastructure, education, and institutional stability, rather than directly engineering the poverty outcomes. The 2013 to 2020 campaign was a direct intervention. The broader reduction from 1978 to 2013, which is most of the achievement, was a byproduct of growth that several political systems facilitated. The celebrated result may say as much about Asian convergence, and about what the global economy did between 1980 and 2020, as about a uniquely Chinese machine. The Party deserves credit, the chapter says, for not obstructing the process, which is more than some governments managed. The credit for engineering the outcome, as distinct from benefiting from the conditions, is harder to assign than the official narrative implies.

Huang's factories had no inspector. Duan has not been publicly explained since 2017. Inner Mongolia's industrial output had been inflated 40 percent. Wukong sold 20 million copies inside a censored market. The 2012 riots burned Japanese dealerships and achieved no diplomatic aim. Li's engine and Milanovic's trade networks sit on the same decades and do not tell the same story.