Saturday, 13 November 1999. A quarter past three in the morning, Beijing.
In the Palace Hotel the telephone rings in the American delegation’s rooms, with word that Premier Zhu Rongji will see them.
The call comes at the end of a bad three days. The American team has landed on the 10th to close the last of the talks that would carry China into the World Trade Organization, and the talks have gone nowhere. Twice the Americans pack and send their suitcases toward the airport, once after Friday’s talks and again after Sunday’s. A White House official tells the hotel’s front desk that the party will leave in the morning. The clerk disagrees. The reservation, he says, is for six days. TIME reports that the Chinese officials who made the booking seem to have known how long the last concessions would take.
A few hours later the Americans are at Zhongnanhai, the leadership compound, in front of the premier.
The scene explains how a deal got made. It does not explain why China was at the table at all, or why its premier would sign away so much. For that the chapter has to go back two years, to the money that left Asia.
The crisis that did not arrive
On 2 July 1997 Thailand’s baht lost about 16 percent of its value in a day. China’s currency held at 8.29 yuan to the dollar, and at 8.28 the year after. By March 1998 the Indonesian rupiah had fallen 74 percent, the Thai baht 37, the Korean won 36, the Philippine peso 33 and the Malaysian ringgit 31. China was not safe because it was strong. Short-term debt was 21.4 percent of its external debt in 1997, against 5.4 percent for India. The crisis reached China as weak demand. Exports were flat in 1998 and growth slipped from 9.3 to 7.9 percent. In February 1998 a Party plenum read the lesson as one for home: the most fundamental thing was to do domestic economic work well.
So Beijing did a great deal of domestic work, fast. In March 1998 Zhu became premier. The State Council plan approved the month before dropped fifteen ministries and commissions, among them electric power, coal, metallurgy, machinery, electronics, chemicals, domestic trade, posts and telecoms, and labour, and cut the number of constituent departments from 40 to 29. At his debut as premier he pledged to reform the indebted state banks in two years and the loss-making state firms in three. The firms were cut as promised. The banks were harder: NPR’s obituary says his efforts on bank bad debt and pensions floundered. The two pledges show the shape of the whole period. Where the state could simply stop doing something, it moved at once. Where it had to build something to replace it, it moved less well. Four of the choices that followed are the subject of this chapter.
The first was to shrink the state sector. In September 1997 the 15th Party Congress, with 2,048 delegates, wrote Deng Xiaoping Theory into the Party constitution. It said public ownership was not the same as state ownership, and that a smaller state share would not change China’s socialist character. Jiang Zemin’s report gave state firms two slogans: grasp the large, let go of the small, and cut workers to raise efficiency. In December Zhu spoke more starkly still, and in May 1998 a State Council and Party conference took up the protection and re-employment of laid-off workers. William Hurst puts the number of laid-off state workers at the end of 1998 at no fewer than thirty million. He also asks whether thirty million jobs were lost in two years or whether they were, in part, only now being counted. Registered unemployment never rose in step, because only the workers of failed firms could register.
The industrial census gives the longer figures. Some 38,100 state firms, employing 15.1 million people in 1998, had exited by 2007. The survivors cut 3.6 million jobs, about 40 percent of what they had. The state’s share of industrial output fell from 50 percent in 1998 to 30 percent in 2005. Hsieh and Song credit the reforms with about a fifth of China’s aggregate productivity growth in 1998–2007. That is the credit side. The debit side, in the NPR obituary of Zhu in August 2026, is about thirty million jobless state workers, protests, and a north-east that never fully recovered.
The second choice was a house. In July 1998 the State Council issued Circular No. 23 on deepening urban housing reform. It ended the allocation of flats by the work unit, sold existing units to tenants at subsidised prices, set up a tiered supply of low-rent, affordable and ordinary commodity housing, and loosened mortgages. Researchers at the National University of Singapore describe it as a step to lift domestic demand after the Asian crisis. It was building on a head start. By the end of June 1998 the housing provident fund held 98 billion yuan, and owner-occupied homes were already more than half of urban housing.
The third was a classroom. In early 1999 the Ministry of Education announced a 47 percent jump in university places, partly in answer to the slowdown and to rising youth unemployment after the Asian crisis. It was followed by 38 percent in 2000 and 22 percent in 2001. Between 1995 and 1998, enrolment had grown by 4.7 percent a year. In 1998 China enrolled 1.08 million new students. In 2008 it enrolled 6.08 million. The share of senior-high graduates going on to higher education rose from 43.1 percent in 1998 to 60.7 percent in 1999. Measured against the age group, China’s tertiary gross enrolment ratio was 6.6 percent in 1999 and 16.9 percent in 2004. India’s was 9.9 percent in 2000 and 11.4 percent in 2004, and by 2004 China was well above it. Three researchers, Li Shi, John Whalley and Xing Chunbing, found that the post-1998 expansion sharply raised unemployment among young graduates, and other studies report oversupply and lower quality at some institutions.
To see what the 47 percent meant, look at the last time a door like this opened. In June 1966 the Party ordered higher-education enrolment postponed, and in July said entrance examinations would be abolished in favour of recommendation and selection. Universities reopened in 1970, but only for people recommended by their work units.
On 10 December 1977, after eleven years, the national college entrance examination was held again, in Beijing on the tenth and eleventh. About 5.7 million people sat it. They were between 15 and 36 years old, and many had spent a decade as sent-down youth in the countryside. About 273,000 were admitted, some 4.8 percent: one candidate in about 21.
From one in 21 to the half of senior-high graduates in 1999 is the distance this chapter covers.
A desk in Washington, a drawer in Beijing
The WTO was the fourth measure, and it was not China’s alone to make: it needed an American signature and, later, an American vote.
In April 1999 Zhu made his first official visit to the United States as premier. He met Clinton at the White House on 8 April carrying concessions meant to close the bilateral part of the talks. According to Susan Shirk, the State Department official who accompanied him, a cabinet meeting the day before he arrived ended with Clinton deciding not to sign. Political advisers argued that signing early would look like conceding too much. Madeleine Albright and Sandy Berger warned that the deal could vanish. At an informal dinner Clinton told Zhu that signing would be delayed. Zhu replied, in Shirk’s recollection, that if that was the way it was, that was the way it was, and later said publicly that he was disappointed. Shirk says Clinton realised almost at once that it was a huge mistake.
On 7 May NATO bombed the Chinese embassy in Belgrade. Three days later China suspended senior military visits and halted its cooperation with the United States on human rights and non-proliferation. Talks resumed on 6 September, by the account of the Shanghai scholar Wang Guoxing.
TIME reports that the resistance inside China centred on what it called “two nevers”: foreign control of telecommunications, and of audio and video entertainment. Clinton telephoned Jiang Zemin on 6 November to remind him that China would be shut out of the coming Seattle round unless it joined that month. The American team landed in Beijing on 10 November.
Then came the six days of the hotel booking. After the call at 3.15 on the Saturday morning, Sunday went badly. At about three on the Monday morning, 15 November, the Chinese side summoned the Americans to the Trade Ministry. Charlene Barshefsky, the American trade representative, answered the Chinese complaints sharply. While she conferred with Gene Sperling, the economic adviser, in a hallway, State Councillor Wu Yi told her that Zhu had come to the building. Zhu, Barshefsky and Sperling settled the last points. Wang Guoxing puts the final stretch in Beijing at 54 hours, and quotes Zhu saying he was criticised unanimously.
What China conceded was a good deal. The package, as TIME listed it, was phased tariff cuts on wheat, chemicals, computers, cars and steel; an end to the state middlemen who held sole rights to import, export and distribute; foreign films raised from 40 a year to 50; retail banking for foreign banks by 2005; an opening of insurance; and foreign ownership of telecoms of up to 50 percent two years after accession. TIME reports that Chinese reformers wanted outside competition to reshape the economy, and that membership could speed reform at home. It does not say what Zhu himself weighed.
The deal was only the first lock. Congress had to remove the annual vote on China’s trade status. The House passed the bill on 24 May 2000 by 237 to 197, and Clinton signed it on 10 October. The law made the new status permanent only once China had joined the WTO on terms at least as good as those agreed on 15 November 1999. It added a 12-year safeguard against surges of Chinese imports and a commission to watch human rights and the rule of law.
Negotiations on the terms were concluded on 17 September 2001. Ministers approved the accession by consensus at Doha on 10 November, China notified its acceptance the next day, and it became the 143rd member on 11 December 2001. It had applied to the GATT, the WTO’s predecessor, on 10 July 1986. The terms were stiff and long. They included an annual review of China’s compliance for eight years, a China-specific safeguard for 12 years and a textile safeguard to the end of 2008. They cut industrial tariffs from a 1997 average of 25 percent to 7 percent, farm tariffs from 31 percent to 14, and car tariffs from 80–100 percent to 25 percent by 1 July 2006. In 2001 China’s merchandise exports were $266.1 billion, against India’s $43.4 billion, and net foreign direct investment inflows were $47.1 billion, against India’s $5.1 billion. Wang Guoxing says China revised more than 2,300 central laws and regulations and rectified some 190,000 local policies. He also lists the costs: losses in farming such as soybeans, collapsed state firms, especially in the north-east, pollution, lost farmland and a wider gap between rich and poor.
The permanence was the point. Handley and Limão estimate that ending the uncertainty over US tariffs explains between 22 and 30 percent of Chinese exports to the United States after accession, on their abstract’s figure. Pierce and Schott found that American industries that had the biggest fall in the threat of tariff rises saw larger import growth from China and bigger job losses after 2001. Acemoglu and his co-authors count 2.0 to 2.4 million net American jobs lost to Chinese import competition over 1999–2011.
On 22 March 2018 the US Trade Representative published its Section 301 findings on China’s technology-transfer practices. The access the 2000 law had made permanent was now the subject of an American complaint. China’s State Council answered in September 2018 with a white paper that began from the premise that some trade friction was only natural. Eighteen years after the vote, the certainty that had pulled factories to China was in question.
The fuses
The chapter’s three domestic measures and its one external one differ in one respect: the state-sector cut was a decision that ended; the other three had no end. Each was a response to a downturn, not a plan for a decade. Housing, universities and the WTO were lit in 1998–99. They are the three fuses.
The university fuse reached the news in 2023. China’s urban unemployment rate for 16-to-24-year-olds rose to 21.3 percent in June. A record 11.6 million graduates came onto the market that year. On 15 August the statistics bureau said it would stop publishing the age-group series. Its spokesman, Fu Linghui, pointed out that more than 65 million of the 96 million urban 16-to-24-year-olds were students, and said it needed further study whether students looking for work before graduation should count.
The housing fuse had a name. In 2020 Beijing introduced limits on developers’ borrowing, the “three red lines”, to cool a sector it judged overheated. Housing and construction made up about a quarter of the economy, NPR reported in 2024. On 29 January 2024 a Hong Kong court ordered the liquidation of Evergrande, which owed about $300 billion. On 20 August 2026 a court in Shenzhen sentenced its founder, Xu Jiayin, to life in prison, finding that between 2016 and 2021 the company had inflated assets and concealed liabilities.
The trade fuse is the one that ran to India. Its cargo flights of April 2025 belong to A Cargo Jet. Apple assembled roughly 55 million iPhones in India in 2025, about a quarter of its output, on Bloomberg estimates reported in India. Three of the five assembly sites are run by Tata and two by Foxconn.
The deadline nobody set
Put the measures in a row and the pattern is plain. A government facing a fall in demand reached for the levers within reach: houses, which could be sold to the people already living in them; classrooms, which could hold young people out of a weak labour market; a treaty, which could bind the state firms to a deadline from outside. In 2001 China’s exports reached $266.1 billion and growth held at 8.3 percent.
The sources cited here say what each measure was for. None of them, as cited, gives an end date for the housing and university measures. The one dated piece is the 12-year American safeguard in the 2000 law. Circular No. 23 and the 47 percent rise were framed as answers to a crisis. What outlasted the crisis was the demand each had created: for flats, for places, for market access. By the time Evergrande fell, housing and construction were about a quarter of the economy. By 2023 a record 11.6 million graduates met a labour market that could not take them. By 2018 the access was the subject of an American Section 301 finding.
The book does not say that any of this was a mistake in 1998. It says that a measure with no end date has no exit, and that the sources cited here record no one, in 1998 or after six days in a hotel in November 1999, writing down what the exit would be.
India’s next decade had no crisis of that size to answer. The next chapter asks how its reforms fared in ordinary years, and finds that the ones that lasted came with a way to be repaired.