Saturday, 23 July 2011. Thirty-four minutes past eight in the evening, on a viaduct outside Wenzhou.
A high-speed train, the D301, has stopped. A storm is overhead; the report would later find that a failed fuse in the signalling had left the line showing clear. Behind it a second train, the D3115, is still moving. At 8.34 it runs into the first, and four carriages are thrown off the viaduct.
One minute later a passenger in the last carriage of the first train types a post on Sina Weibo. The train has stopped suddenly, she writes; she felt a strong collision; the power is off; she hopes it is nothing serious. Her post goes out before any report from state media, which The Hindu’s correspondent in Beijing called a first for a major accident in China.
By Sunday the crash has been mentioned more than four million times on the site. A historian’s post predicting that someone would be praised, a scapegoat punished and the government would carry on as before is forwarded more than 60,000 times. The railways ministry is criticised for burying the wrecked carriages at the site. Forty people died.
The question this chapter asks is what the viaduct was made of. Not the concrete: the money.
2008BeijingOn the evening of Sunday 9 November 2008 the State Council said the global financial crisis was intensifying, and put a number on its answer. The plan was 4 trillion yuan, about $586 billion, over two years, roughly 7 percent of GDP a year, for railways, subways, airports, low-income housing, rural roads and the rebuilding of the earthquake zone. In March 2009 the planning commission gave the split: 1.5 trillion for infrastructure, 1 trillion for reconstruction, 370 billion each for rural development and technology, 210 billion for the environment and 150 billion for education and culture. The centre was reported to pay 1.18 trillion of it. The lines listed there add to 3.6 trillion; the article does not give an amount for the housing and welfare lines.
That is less than a third from the centre. The economists Chong-En Bai, Chang-Tai Hsieh and Zheng Song, who reconstructed the episode for the National Bureau of Economic Research, put the whole programme at about 12 percent of a year’s GDP (estimates of the size differ with what is counted) and say it was largely financed by off-balance-sheet local financing vehicles, which went on spending about 10 percent of GDP a year after 2010. Behind that lies a rule. China’s 1994 budget law barred local governments from borrowing directly. In March 2009 the banking regulator said it would encourage local governments to attract bank lending to centrally set investment projects, including through government investment and financing platforms, which avoided a technical breach of the law. Local governments set up companies to borrow and spend on their behalf. Gross capital formation rose from 42.0 percent of GDP in 2008 to 44.9 percent in 2009.
It worked, on its own terms. Growth was 9.7 percent in 2008 and 9.4 percent in 2009. And some of what was built was good. The railway plan of 2004 had been revised in 2008 to 16,000 kilometres of passenger line by 2020; the first 350 km/h line, Beijing to Tianjin, opened on 1 August 2008 and carried more than 16 million passengers in its first full year. The World Bank later reckoned that Chinese lines cost $17 to $21 million a kilometre, about two-thirds of the cost elsewhere. The train that crashed at Wenzhou ran on that system. The stimulus that expanded it was largely paid for outside the main accounts.
2008DelhiIn India the same two years looked different, and the difference is the point of putting them side by side.
On 22 July 2008 the Lok Sabha was packed, because both sides had summoned members from sick beds and, in some cases, from prison cells. About fifty votes were cast on paper, which delayed the count. The government’s Left allies had withdrawn support over the civil nuclear agreement with the United States. During the debate, BJP members waved fistfuls of cash and alleged they had been offered bribes to abstain. Hours later the Speaker announced 275 votes to 256.
The crisis response that year was large, and it was on the Centre’s books. The Sixth Pay Commission award cost 0.65 percent of GDP, the farm-loan waiver 0.3 percent and higher food and fertiliser subsidies 1.03 percent. With a stimulus of nearly 3.5 percent of GDP on top, the Centre’s fiscal deficit went from 2.7 percent of GDP in 2007-08 to 6.2 percent in 2008-09. The Reserve Bank cut its policy rate from 9 to 5 percent. Growth fell from an average of 8.8 percent in 2003-04 to 2007-08 to 6.7 percent in 2008-09, on the figures of the day. Inflation, by the World Bank’s series, ran at 10.9 percent in 2009 and 12.0 percent in 2010. Recovery was quick: by the World Bank’s series India grew 7.9 percent in 2009 and 8.5 percent in 2010. India paid on the Centre’s books, and in prices.
India’s counterpart was in the banks. It surfaced in 2015, when the Economic Survey counted how much corporate debt sat with firms whose earnings could not cover their interest. That is the bill that comes later in this chapter.
The Games were the part everyone could see. On 4 August 2010, two months before the Games, the Games Village was unfinished. The budget approved in April 2007 had been £500 million, and press estimates now ran from £2 billion to nearly £4 billion. Campaigners said nearly a hundred construction workers had died on the sites. The Guardian’s reporter noted that the Beijing Olympics, two years earlier, had been widely seen as a success. The Delhi overrun was argued over in courts, in audits and in the press. The first Chinese count came a year later, and the full one three years later.
2011BeijingAfter Wenzhou the state did three things in order. In December 2011 the State Council’s investigation reported serious design flaws in the signalling and control equipment and sloppy management, and named the former railways minister, Liu Zhijun, who had been sacked in February over corruption allegations, as bearing main leadership responsibility. It criticised 54 officials. New high-speed projects were halted for a time and speed limits imposed, and the ministry planned 400 billion yuan of rail investment for 2012, less than the year before. Then the building resumed. The network was 16,000 kilometres at the end of 2014, six years ahead of the 2020 target, and more than 19,000 at the end of 2015, over 60 percent of the world’s total by the government’s count.
That is the story of the train. The story of the money was being counted at the same time.
On 27 June 2011, less than four weeks before the crash, the National Audit Office reported that local governments owed 10.7 trillion yuan, about $1.6 trillion, at the end of 2010. Reuters Breakingviews judged the audit’s figure to be on the low side. It noted that more than half the loans in some sectors, motorways among them, had simply been rolled over. The State Council had warned of the risk the year before, in Document 19 of June 2010, and ordered a clean-up of platform debt.
On 30 December 2013 a second audit put local debt and contingent liabilities at about 17.9 trillion yuan at the end of June 2013, up 67 percent from end-2010. It was the first time total government debt figures had been released. The same report quotes an economist, Lu Ting of Bank of America Merrill Lynch, warning of rapidly rising leverage but arguing that China was not on the brink of a crisis: central-government debt was 21 percent of GDP, the debt was in yuan and owed at home, and reserves were $3.5 trillion.
On 19 July 1969 the Indian government nationalised fourteen banks, those with deposits above Rs 50 crore. Bank offices rose from 8,045 in June 1969 to 65,931 in June 2001; by then public-sector banks held 60,531 of them.
Forty years after the nationalisation the economist Shawn Cole tested whether government-owned banks lent to the political map. He found that government-owned banks’ agricultural lending rose by 5 to 10 percentage points in election years, and that the targeting followed the political map.
The two stories rhyme on one point. In China the 1994 budget law barred local governments from borrowing directly, and the borrowing came back through companies. In India the state owned most of the lenders, and one measured result is that farm lending rose in election years.
State Council Document 43, dated 21 September 2014 and published on 2 October, changed the instrument rather than repeating the warning. Provincial governments could borrow moderately, and city and county borrowing had to go through the province. Local borrowing was to take the form of bonds. Financing platform companies were stripped of their government-financing function and barred from adding government debt, and borrowing was brought under national quotas and full budget management. Projects under construction were to be allowed to finish.
It was four years from the clean-up order of Document 19 to the change of instrument, with two audits between them. Credit did not stop. By the Bank for International Settlements’ figures China’s total non-financial credit stood at 170.8 percent of GDP at the end of 2010 and 238.6 percent at the end of 2016, of which general government was 33.2 and then 50.1 percent.
2024On 8 November 2024 the Standing Committee of the National People’s Congress approved a 6-trillion-yuan rise in the ceiling for local special bonds, to replace hidden debt, at 2 trillion a year through 2026. The finance minister, Lan Fo’an, added that another 4 trillion of special bonds could be used for swaps over five years: a 10-trillion-yuan package in all. Hidden debt, which the ministry had identified at 14.3 trillion yuan at the end of 2023, would fall to 2.3 trillion by 2028.
Do not read this as the 2008 debt coming home. Document 43 had barred platforms from adding government debt ten years earlier, so the 14.3 trillion is a count of liabilities that survived that ban or were built after it. The sources here do not say how much is which, and the Chinese report of the swap does not date the debt. What can be said is that ten years after the order to close the side door, the liabilities it produced were large enough that in 2024 the centre took them onto its own balance sheet.
India’s version of the count came in 2015. The Economic Survey of 2015-16 reported that the share of corporate debt owed by stressed companies, whose earnings could not cover their interest, had risen from 35 percent in December 2014 to 41 percent in December 2015, and that public-sector banks’ return on assets was less than a third of the 1 percent norm. It described stress in banks and firms together and set out four remedies: recognition, recapitalisation, resolution and reform. The government had launched a phased recapitalisation programme, Indradhanush, in August 2015, and the Reserve Bank had begun a review of banks’ asset quality.
The Insolvency and Bankruptcy Code was introduced on 21 December 2015. A joint committee reported on 28 April 2016, the Lok Sabha passed it on 5 May and the Rajya Sabha on 11 May, less than five months after introduction. Its core was a 180-day clock for resolving a company’s insolvency.
2025In December 2025 Parliament’s Standing Committee on Finance reported that the average time taken to close a corporate insolvency resolution process was 713 days, against a mandated 330. It blamed a shortage of tribunal benches, vacant posts and litigation. The law had set a clock. The tribunals that were meant to keep it were short of benches and staff.
By the end of 2014 China’s reserves were $3.90 trillion, the highest year-end level in the World Bank’s 2009-17 series. Gross fixed capital formation was 43.4 percent of GDP. The high-speed network had reached 16,000 kilometres. The same government work report that gave the rail figure said local-government debt management had been strengthened. From 2015 the reserves began to fall.
The Indian Economic Survey of 2016-17 drew the comparison itself. It put India’s stressed loans, meaning non-performing, restructured and unrecognised, at about $191 billion in September 2016, 8.4 percent of GDP and 16.6 percent of loans. For China in 2015 it gave $1.3 trillion, 12.0 percent of GDP and 15.5 percent of loans. India’s path, it observed, had resembled China’s on a smaller scale, with banks providing generous finance to keep over-indebted firms alive. The difference it left standing was that India’s accounts were published, argued over and addressed by a law, while China’s number had to be found by an auditor.
Between them the two countries show the two ways a boom’s bill arrives. China’s came as a number that had to be found, 10.7 trillion in 2011 and 17.9 trillion in 2013, because the vehicles that borrowed it had no line in the budget. India’s came as a ratio, 41 percent of corporate debt in stress in 2015, because the lenders it had told to lend were the same lenders that held the losses. In China the centre eventually took the liability onto its own books. In India the Budget speech of 2018 set out a recapitalisation plan of Rs 2.11 lakh crore, of which Rs 80,000 crore was in bonds in 2017-18, and the courts were left to do the rest.
Neither is a story of incompetence. A state that could lay 19,000 kilometres of fast rail and a state that could pass a bankruptcy code in five months were both showing a capacity. What China lacked in 2009 was a lane in which local investment borrowing was visible and capped; Document 43 built one in 2014. India’s borrowing was visible, but it was the lenders’ losses that were not.
One Indian liability was hidden neither in a platform nor in a bank. It sat with the state utilities that buy power and sell it on, and in July 2012 it showed itself not as a number in an audit but as a grid that went dark.