US, CHINA AND TRADE · 12 SEPTEMBER 2026 · 13 MIN READ

Underwriting the Chinese Economic Model

32 days in China, 15 cities and a very long, fascinating road trip - numerous first-hand experiences and so many conversations with entrepreneurs, factories, academics and new friends, leading to some reflections.

I have talked about Shenzhen and Guangdong. There is a lot more to say.

By the end of 2025, China’s high-speed rail network passed 50,000 kilometres. That is more than all other countries combined. It moves 4.28 billion passenger trips a year. It reaches 97% of cities with over half a million people. Shanghai became the first port in the world to handle 50 million TEU. Ningbo-Zhoushan has led the world in cargo tonnage for sixteen years running. In 2025, CATL and BYD produced 55.6% of the world’s EV batteries. Robot density reached 166 units per 10,000 factory workers, up 17% in a year. That makes China’s the largest industrial robot fleet on earth - about 4.5 times Japan’s. In 2025 alone, China added over 430 gigawatts of wind and solar power. In May, it installed about as much solar as Germany’s entire base. Manufacturing value-added remains at 28-30% of the global total. That is more than the US, Japan, and Germany combined. Thinking of China as just a source of cheap plastic toys is an outdated notion.

And there is a clear, delightful human element too. By 2025, urbanisation reached about 68%, up from 36% in 2000. As per the World Bank, China has lifted nearly 800 million people out of extreme poverty since 1978. That is about three-quarters of all global poverty reduction. Basic pension coverage now reaches 1.07 billion people. Basic medical insurance covers around 1.33 billion - about 95% of the population. Rural incomes are slowly catching up to urban incomes, with the gap narrowing to about 2.34 to 1.

And as always, there are also elements that ground you.

One afternoon, my phone suddenly lost all connection to the outside world. No warning, just gone. It came back a little over an hour later. I found out that China intentionally shut off all HTTPS traffic to overseas sites during that time, for routine maintenance. No one around me seemed surprised. This is just normal here. That moment stayed with me more than any GDP figure I’ve shared before. It’s what I want to focus on now.

Because that’s the real story here. Not just what China built. What holds it up. And what it costs to keep it standing.

Three things carry that cost. The hukou system splits the country in two. The information wall keeps you from checking what’s true. And underneath both sits a pile of debt that nobody outside a few rooms in Beijing can fully see.

China’s somewhat outdated hukou household registration system still divides the country into two classes, on purpose. The system keeps a large, cheap, and flexible labour force moving into coastal factories, without giving them full urban rights. In 2022, only about 47.7% of people in cities had an urban hukou, even though 65% of the population lived in cities. That means 200 to 300 million people live in cities as second-class residents. They work and pay taxes there. But they’re often shut out of good schools, healthcare, and housing rights. Some reforms have helped - over 40 million rural migrants got urban hukou between 2021 and 2023. But there is a long way to go.

In my view, this is one key reason household consumption in China remains about 38-40% of GDP, compared to over 60% in OECD countries. Everyone I spoke with, from policy analysts to businesspeople, agrees the economy needs to lift domestic demand. But almost no one with the power to change things wants to pay for the safety net that would let families spend more instead of saving out of fear. This isn’t just a cultural habit of saving. It’s a logical reaction to a system that hasn’t provided the kind of insurance modern consumer economies need. People save because they know the state won’t cover their risks.

The Great Firewall was, in its own way, the biggest inconvenience of the trip. Like most foreigners, I had trouble connecting to my usual systems from home - a new experience for me. I am so used to the internet and cloud systems being ubiquitous that not being able to reach global hotel and airline websites was a shock to the system. Let alone news or social media sites. The firewall extends to basic financial transactions too. Using credit cards or Apple Pay is a bit of an undertaking. Alipay and WeChat reign supreme, with a healthy clip on the ticket on every transaction.

In 2025, Freedom House scored China 9 out of 100 for internet freedom. That makes it one of the most restricted places online. The Ministry of State Security has made using VPNs to bypass these blocks illegal, and enforcement is quite strict. Under the new Counter-Espionage law, “espionage” includes broad terms such as access to “documents, data, materials, and items related to national security and interests.” The law has been enforced in high-profile cases. Mintz Group staff were detained and the firm fined about $1.5 million. Bain’s Shanghai office was raided, devices seized. Capvision was investigated on state TV. The message is clear. Now, even basic due diligence - checking whether numbers are accurate - can be labelled spying, if the authorities choose to. By criminalising this kind of checking, the government doesn’t just scare off journalists. It makes foreign investors factor in risks they can’t verify. And when investors can’t check things for themselves, they often decide to leave.

The political crackdown on private capital in 2020-2021 caused real and lasting harm to the entrepreneurial energy that drove China’s coastal growth. The efforts to win back confidence since then haven’t repaired the damage.

Ant Group’s $313 billion IPO was cancelled in November 2020. Didi had to delist and was banned from signing up new users. In 2021, the government effectively nationalised private tutoring by order. The “Common Prosperity” campaign brought a wave of billion-dollar “donations” from nervous tech founders, along with talk of controlling the “disorderly expansion of capital.” Then, in February 2025, President Xi Jinping met with these entrepreneurs - including a rehabilitated Jack Ma - and exhorted them to “showcase their talents.” In the AI era, Beijing needs private-sector jobs, technology, and confidence much more than it did in 2021, and it has said so openly.

The response has been lukewarm for good reasons. After two years of showing founders that success can bring trouble, it’s no surprise they’re cautious now. The numbers back this up. In 2025, China ranked second in the world for high-net-worth outflows, with a projected net loss of 7,800 millionaires. Utilised FDI dropped 27% to about $115 billion in 2024. Balance-of-payments net FDI fell to around $18.6 billion - a three-decade low, down from about $344 billion in 2021. The EU Chamber’s 2025 survey found a record 73% of members said things had gotten harder year on year, and profit optimism hit a record low of 12%. AmCham members also reported a record number considering relocation.

But I also found that the Western media’s narrative of a draconian Social Credit system was way off the mark. There isn’t a single, dystopian citizen score. Instead, there’s a mix of national blacklists and a much larger Corporate Social Credit System, one that handles about 37 trillion yuan in compliance-linked financing. It’s less like a sci-fi show and more like a big administrative tool that makes up for weak courts. Less dramatic than the myth. But actually more important for businesses, since it affects your access to financing, not your social media activity.

Here is where the three threads meet. The sprawling capital infrastructure gives you a glimpse into how the governance logic that suppresses information and disciplines private capital actually gets financed. And the nation’s debt picture is grave.

Chinese debt isn’t split neatly into sovereign, corporate, and household buckets the way most market economies do. To get a clearer picture- take the pure sovereign debt running at 55-60% of GDP, add the quasi-fiscal layer of local financing vehicles and policy-bank bonds, which add another 48-52%. That pushes the IMF’s own “augmented” government debt estimate to somewhere around 110-124% of GDP. Now add state-owned enterprises (SOEs), which carry a further 115-120%. Add it all up, and you land on a consolidated public and quasi-sovereign perimeter of roughly 230-250% of GDP. That’s the real size of the government’s balance sheet, once you stop pretending the state and its enterprises are separate things - which, in the case of China, are one and the same.

The Foshan pivot I flagged last time - the robotics funds, the humanoid line - is a good example of this constraint, not an exception to it.

Guangdong was the first province to declare “zero hidden government debt,” back in 2022. It has held that status since then- one of only four jurisdictions in that position by early 2026. As a direct result, it got left out of the national debt-swap programme. It was told, in effect, to clean up its own mess before asking Beijing for a hand. And it did.

China’s showcase province earned its strength by absorbing its own liabilities first. Which means it’s got no slack left for the next downturn. The strain shows unevenly even inside the province. Shenzhen is genuinely resilient. But Guangzhou and Foshan, more exposed to property and old-school manufacturing, are visibly slowing. Both cut mortgage rates toward 3% in 2024 to prop up housing. Empirical evidence points to real pay cuts for police in the province, and wage arrears for factory and state workers in Dongguan.

Once you move inland, it sharpens further. A peer-reviewed Princeton study found Chongqing’s local financing-vehicle debt hit 915% of fiscal revenue - among the highest of any major jurisdiction in the country. In August 2024, officials in Chongqing’s Bishan district said it was time to “smash pots and sell iron” - an idiom for liquidating everything you’ve got. They stood up a task force with that exact name. Chongqing raised nearly 200 billion yuan from state-asset sales in 2025, and is chasing the same again in 2026. Its official special-debt outstanding has roughly tenfolded over the past decade.

Sichuan, home to Chengdu, was one of the biggest recipients of the central debt-swap allocation. It replaced over 100 billion yuan of hidden debt through the refinancing programme. Support Guangdong never got, because it never had to ask.

The common thread is land. Revenue from land-use-right sales fell 14.7% nationally in 2025, to 4.15 trillion yuan. That’s the fourth straight year of double-digit decline, more than halved since the 2021 peak. Residential land sales specifically are down roughly 65% from their 2020 high. Land and property taxes used to give local governments about 36% of their revenue. Now they give around 13%. That’s a big gap, and a hard one to fill, because it was never really a financing problem to start with. It was a revenue model that’s simply run dry.

Which puts the robotics push I described last time in a different light. Foshan’s pivot runs on at least 3.5 billion yuan across two industrial funds, plus a separate billion-yuan innovation fund, plus tens of millions a year in R&D grants. That’s part of a national web of some 1,800-plus government guidance funds. One Boston University research centre says these are explicitly built to pull in bank, insurer and state-enterprise capital “while avoiding direct fiscal transfers.” That phrase says the whole thing in miniature.

The IMF puts the fiscal-equivalent cost of China’s industrial policy at 4.4% of GDP as of 2023 - cash subsidies, tax breaks, cheap land, cheap credit. It warns bluntly that this is crowding out household support, while locking in overcapacity in the very sectors it’s meant to grow. CATL’s disclosed government subsidies went from $77 million in 2018 to $809 million in 2023. Chongqing’s Changan has committed 100 billion yuan through 2028 to its EV brands, funded partly through the same municipal and national green-development vehicles already carrying the debt I’ve just laid out.

The recapitalised banks doing the lending are the same banks in trouble. Beijing pumped 500 billion yuan of special sovereign bonds into the big state banks in 2025, with another 300 billion pencilled in for 2026. Yet their net interest margins hit a record-low 1.42% sector-wide - squeezed by exactly the directed lending to stressed local financing vehicles and property developers that the recapitalisation was supposed to fix.

Beijing’s headline fix - the roughly 10-12 trillion yuan local debt-swap programme launched late 2024 - is running on schedule. It’s doing exactly what it was built to do. Officially recognised hidden debt has fallen from about 14.3 trillion yuan at the end of 2023, to something closer to 7.4 trillion by late 2025. The number of local financing vehicles has dropped 71% since March 2023.

What it’s not doing is close the underlying revenue gap. S&P calls it a temporary fix - buying time through roughly 2027, without repairing the cash flows underneath it. Fitch reckons the restructured debt covers only about a quarter of the hidden total. The Atlantic Council’s term for it - extend and pretend - is the right one. It turns opaque, high-interest shadow debt into transparent, low-interest sovereign debt. It doesn’t make the debt smaller. It just puts it on a longer clock.

How big that problem actually is stays genuinely contested. It sits in the same information fog running through everything else. When the IMF’s 2024 assessment cited roughly 44 trillion yuan in local financing-vehicle debt - three times the People’s Bank governor’s own figure of about 14.8 trillion - Beijing’s response was to dispute the IMF’s data source, rather than publish a number that reconciled the two. Foreign analysts have been locked out of the Wind financial database, one of the few tools that can aggregate this independently, since May 2023.

The gap between what the state says its balance sheet looks like, and what independent institutions think it actually looks like, isn’t a rounding error. Put the two together, and the  scenario isn’t rosy. The IMF’s latest Article IV assessment has growth decelerating to about 4.5% in 2026, and, without real reform, to roughly 3.5% by 2030. The GDP deflator stays negative into 2026 - a textbook, if managed, debt-deflation dynamic. Independent forecasters at BOFIT see growth closer to 3% by 2028. Fitch downgraded China’s sovereign rating in April 2025 on rising debt and deficits, projecting explicit government debt near 74% of GDP by 2026.

No local financing vehicle has defaulted in the public bond market. The containment is real, and genuinely impressive as a piece of financial engineering. But commercial-bill delinquencies and payables defaults, while still comparatively rare, rose materially through 2023, and have kept climbing since. Independent economists estimate unpaid contractor fees and civil-servant wages nationwide somewhere between 4.5 and 10 trillion yuan.

This is a slow, managed grind of a system stretching every maturity it can, for as long as it can. Because the alternative—genuine restructuring, or an actual transfer of resources from the state to households—keeps getting knocked back.

In my view, one needs to separate the supply story from the demand story in every China thesis. If the exposure is to Chinese manufacturing capability - batteries, solar, EVs, automation - the moat is real, and probably still widening. The state will keep financing it even under fiscal stress, because it’s decided this is the one bet it won’t unwind. But exposure to Chinese domestic consumption requires caution. The rebalancing is happening a point or two of GDP per decade, not per year. Because the debt architecture and the hukou system are both, structurally, pulling against it at the same time.

But perhaps the most important element for foreign investors is the due-diligence tax. Under the current Counter-Espionage regime, the cost of actually knowing what one owns in China has gone up. Without normal verification, one could be carrying an unpriced risk. Either fund a presence that can survive scrutiny, or discount the position for the opacity you can’t resolve. I will be watching FDI inflows actually turn positive year-on-year and high-net-worth outflows drop meaningfully below the 2024-25 pace. This will be the real-time referendum on whether this model holds together. If producer prices stabilise, and the overcapacity sectors consolidate without a demand collapse, Beijing’s supply-side bet is working better than the IMF currently thinks. If deflation digs in, or the trade truce with Washington frays after its November 2026 renewal date, the export engine stalls with no domestic engine built to take over. And I will be keeping an eye on Guangdong specifically. Not because it’s fragile, but because it’s the province with the least room left to absorb the next shock, having already spent its slack proving it didn’t need a bailout.

Fast-growing state-directed economies get written off easily. The write-offs generally have usually been wrong on timing far more often than on substance. Back in 2003 and 2004, touring Japanese and Korean car plants, economists tipped mass labour substitution from robotics on a scale that never turned up on schedule. Productivity gains landed closer to 3% a year than the 75% displacement some were calling. It took three decades, not three years, for the real picture to show. I reckon China’s demand-side rebalancing runs on a similarly long clock – longer than most commentary out of the West has the patience for.

I don’t think China’s growth rate answers that question. I think that’s the question that decides whether this decade’s miracle compounds into the next one or just gets refinanced until it can’t be anymore.

Discuss this analysis. Tararua Economics briefs governments, sovereign institutions and investors on these questions. Contact