Economist Lan Xiaohuan proposes better social security to fix China’s unbalanced economy and weighs the impact of technology on lives now and in the future
Lan Xiaohuan is a professor of economics at China Europe International Business School. His book, How China Works: An Introduction to China’s State-led Economic Development, has sold millions of copies in China and has been translated into multiple languages.
Here, he discusses the economic realities behind China’s record trade surplus, the case for a stronger social safety net, and how public data infrastructure shapes the artificial intelligence race with the United States.
SCMP Plus readers get early access to articles in the Open Questions series.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.
Since you wrote How China Works, the global and domestic environments have both evolved. What are the most critical shifts and developments for China? If you were to add a new chapter to the book to reflect these changes, what would be its central theme?
I would not add a new chapter; I would write an entirely new book. It’s not that there aren’t new issues, but rather addressing them requires a new framework.
Domestically, the decline of the real estate market and the rapid rise of manufacturing and technology are significant shifts. My new book will likely focus on China’s industry and technology.
But actually, I think the most consequential changes are external: the impact of the pandemic, the Russia-Ukraine war, the energy shocks regarding Iran, and shifting political dynamics in the United States. In comparison, China’s domestic baseline has remained relatively stable.
How China Works explains to the world how to view China. I’ve also thought about writing a book explaining to Chinese people how to view the world. Historically, global world views have been dominated by a North Atlantic, Euro-American perspective. So what is the world from a Chinese perspective? In terms of economics, that would mean looking more closely at Southeast Asia, Central Asia and Russia, where we share closer relations but don’t yet have a systematic understanding of them.
Amid a property market downturn, how can local governments secure new, stable streams of income?
Fiscal pressure can’t be solved by money alone, nor by changing the development model alone; it needs both. Current policy adjustments – such as relaxing special local government bond quotas and easing bank credit – help refinance old debts with new ones to prevent defaults. This is meant to help buy time.
Long-term fiscal sustainability is not only about increasing income – it requires aligning expenditure with income. Past local government spending was too high. It’s unrealistic to expect land revenue to support such a large amount of spending, or to expect other revenue sources to plug the fiscal deficit left by the property market downturn. There aren’t any long-term sustainable revenue sources that can directly replace land finance.
Therefore, I think a reasonable approach is tax reforms in the long term, alongside spending cuts. Much of China’s core infrastructure – high-speed rail networks, urban undergrounds, major city expansions, and so on – has already been completed, so these kinds of local government expenditures will naturally decline.
Local governments might still want to launch new projects, such as building airports, but their finances might not allow it. Investment remains a major engine of economic growth, but its share is not as large as before, nor should it be.
For tax reforms, perhaps it’s not exactly about increasing local government revenue, but about making the distribution of the tax burden more reasonable, such as through adjustments to the value-added tax and consumption tax. But how exactly should they be adjusted? Even a fraction of a percentage point difference in tax rates carries major consequences. So while scholars have offered various opinions, these proposals mean little until implemented.
You mentioned that investment’s share as a growth driver must decline. Given the slow pace of rebalancing so far, what structural changes are needed for Beijing to boost domestic demand and bring household consumption in line with global benchmarks?
There are two ways to increase income: raise wages, or raise property and asset income. In terms of wages, most citizens do not receive income from their labour from the state; they are employed by private enterprises. The state cannot simply mandate private sector wage rises, it can only optimise the business environment in hopes that employment wages improve.
For property and assets, this requires stabilising property values, improving financial asset performance and unlocking the latent value of rural land, such as through asset revitalisation.
The main thing that the government can do directly is to step up social security and strengthen spending on education, healthcare and pension security. China’s current savings rate is relatively high, so hopefully these measures can help reduce the savings rate and increase consumption.
But I hold a slightly different view from many economists on this issue, in that I think public welfare spending should be expanded regardless of whether it boosts consumption. Creating a strong social security system is what the government should do, and it shouldn’t be linked to consumption. Whether that triggers consumer spending is outside direct policy control, as citizens have a right to maintain a high savings rate even with a robust safety net.
The transition to raise consumption is a long-term process. Household consumption currently accounts for around 40 per cent of GDP in China.
In the United States, it accounts for around 70 per cent. But the number is so high because it is driven by housing, healthcare and education expenses. In China, healthcare and education – which form the bulk of service consumption – are not priced according to the market and are pretty affordable. In the US, healthcare and college tuition are relatively expensive. Do we really want to depend on this to drive consumption? I don’t think so.
Germany and Japan are better international benchmarks, with household consumption at around 50 to 55 per cent of their economies. But if China raises its consumption share by one percentage point per year, the rebalancing will still take over a decade.
Weak household consumption and industrial overcapacity have also driven China’s trade surplus to record highs. Beijing has signalled that it wants to rebalance these trade flows, but how successful can it be if demand remains anaemic?
China’s trade surplus issue isn’t just about weak domestic demand and overcapacity – it also requires a willing buyer. China cannot force foreign markets to purchase its goods. One country’s exports and another country’s imports are two sides of the same coin. The surplus arises because foreign markets genuinely need these goods and demand is substantial.
While foreign markets do want to restrict Chinese imports, what they really want is for China to set up factories there to sell the goods directly, while creating jobs and facilitating technology transfers. Trade barriers can’t restrict demand, they can only aim to take a cut of the profits and redistribute them.
From China’s side, imports will surely expand, especially as the appreciation of the yuan makes foreign goods and services cheaper. But whether this expansion is driven by goods or services is still uncertain.
Raising goods imports might be difficult. The things that China wants – such as graphics processing units and lithography equipment – others won’t sell. What else are we supposed to buy? American cars? That is hardly realistic.
Customs data – including the US$1.2 trillion trade surplus – only represents trade in goods and has significant limitations.
Supply chains have become extremely complex. Many goods recorded as imports from Europe, Canada, Mexico and Southeast Asia might actually come from China, and vice versa. Goods that China imports that might not appear to come from the US or the European Union could actually be from these places too.
Importing more services is possible, but it is difficult to calculate. For example, spending money while travelling overseas and sending children abroad to study count as importing services, but customs doesn’t track this. It can only be calculated through foreign exchange data. The current account surplus theoretically includes services, but service data is more complicated in reality. It’s hard to say who holds the trade surplus for services.
China’s services imports have also been constrained by external political factors, such as strict visa policies or restricted access to certain educational programmes in the US. The fact that there are fewer Chinese students studying in the US now, that’s completely caused by the US themselves. Chinese people genuinely want to buy – previously goods, now services – but frankly, imports are severely restricted.
Many people have noted that data has now become a new factor of production, with some comparing it to oil. What’s your view on this?
Data as oil is just one analogy. I think it’s probably outdated now in the age of AI, though many data dispute rulings still work on the logic of this analogy.
But data is different from oil. First, if you extract it, that does not stop me from extracting it too – oil does not work that way. If there is oil under a piece of land, your extraction precludes mine. Data does not have that problem; we can both use it.
Second, oil is a natural resource – either you have it or you do not. Data availability, however, can be generated by policy. The data exists; it is simply a matter of whether you allow access. If you adjust usage policies – as Europe has done by relaxing certain rules – a vast amount of data suddenly becomes accessible. Therefore, the critical factor with data is what the law dictates: how much is recognised, what can be used, and under what restrictions. I think that regulatory boundary is the primary constraint.
Oil is a very tangible resource, so ownership is easy to determine. If it is under my land, the oil belongs to me. Data ownership is not nearly as simple. How do we regulate this? A unified model has yet to emerge globally, and it is difficult to see how one will.
There are other analogies to compare data to, and each one reveals different characteristics. For example, data is also inherently a product of labour. Let’s say you drive an electric vehicle. You’re constantly contributing your data to autonomous driving. The more you drive, the more data you contribute, but you don’t get paid for it – that’s frustrating. You continuously generate data for others to use, but you don’t receive corresponding compensation for the data generated by your actions, while others profit from this data.
How will this unique role of data affect the development of AI across different regions and their economies?
Frankly speaking, when it comes to the development path of artificial intelligence, only China and the United States are relevant right now. Other countries lack foundational models. Europe has a presence but faces real structural headwinds, including strict data rules.
The primary differentiator, then, boils down to who controls the data. In China, many core data assets are public property, meaning a wider variety of market participants can develop applications on top of them. In the US, however, commercial data holders always command a share of AI revenues.
Take the legal sector as an example. In China, legal data repositories are provided directly by the government via the Supreme People’s Court, allowing anyone to access them. In contrast, the US market is dominated by two private data vendors, so data is more expensive to access. The two vendors are also developing their own AI tools, making it difficult for independent developers to enter the space.
Similarly, Chinese municipalities and state-owned enterprises accumulate vast amounts of smart city and intelligent transport data. If held as private property, this data would be commercialised for profit. But if it is treated as basic infrastructure, the data owners do not extract direct profits from it.
From the perspective of benefiting AI development, accessible, high-quality data – not completely free, but requiring just a nominal cost – is definitely more beneficial for applications.
Think about the many companies in China developing autonomous driving systems: a significant reason is the relatively readily available underlying data, thanks to the infrastructure built by the government in the early stages. You don’t just need vehicle data, you also need various other data like road data and traffic light data. It needs to form a complete ecosystem, and I think you can’t do without public data.
How close, then, is China to catching up with the US in the AI and tech race? In which areas is China leading, and where does it remain most vulnerable?
Whether China catches up or surpasses the US in AI development is not important. What matters is which country can better utilise AI and leverage it for its own economy. The US leads in foundational models and computing power. But does this constitute a guaranteed advantage for the industry? Not necessarily. It also depends on electricity availability, data access, as well as the industry to which you apply it and the results it generates.
For instance, the US excels in applications of AI in legal and financial fields. But for manufacturing, there are a few constraints. If a factory has not undergone digital transformation, or if its data is fragmented, practical application remains far off.
AI models are universally applicable; the key is whether your infrastructure is ready – meaning whether your data is structured and your workflows are prepared. This is a new challenge for everyone, but I think China’s manufacturing sector has more application potential than the US.
Cost is also a significant factor when it comes to widespread use for industry applications, and American models are considerably more expensive. Ultimately, the market will find a compromise in terms of cost-effectiveness.
Some have voiced concerns that advanced tech sectors like AI and humanoid robotics are at risk of “involution” in China. What are your thoughts on this, and how can policymakers prevent these strategic industries from falling into the same hyper-competitive trap?
Robotics are pretty competitive right now, but that’s not a risk yet. Humanoid robotics are still a long way off – the GPT moment for robots hasn’t arrived yet, so everyone thinks there’s still opportunities.
Involution happens when the technological path is already set, and everyone rushes in. In the photovoltaic sector, for example, the playbook is clear. Companies expand capacity aggressively along established pathways while making incremental technical refinements. But for humanoid robots, the approach remains undefined. The industry is still in an exploratory phase. There’s no overcapacity now since robots aren’t widely used yet. Having hundreds of firms enter a sector at this stage is more of an exploration rather than involution.
In contrast, I think AI large language models have already reached their limits in terms of competition. The so-called “war of a hundred models” has consolidated because small firms cannot sustain the immense capital expenditures required for AI computing. The market has naturally winnowed down to a handful of players.
Regarding policy guidance, I think these industries are still too young. Policymakers must understand what they are regulating. For instance, policies for the photovoltaic sector are relatively easy to formulate because the crux of the problem is already known. But right now, for sectors like humanoid robots, even some industry insiders cannot yet predict where the technology will land.
While policymakers and entrepreneurs express optimism regarding technology, there is palpable public anxiety around AI, given stagnant wage growth and elevated youth unemployment. How worried should we be?
I think worrying is pointless because we have reached a point of no return. AI is here to stay, and we simply have to accept it. While many economists argue that AI will cause unemployment but simultaneously create new jobs, I am sceptical of how this rebalancing will play out.
There is likely to be significant demand for manual labour – which AI cannot replace for the time being – but displaced white-collar workers are probably unwilling and unable to take on these roles.
And actually, even manual labour faces risks of displacement. I recently spoke with representatives from a major logistics company who shared that the number of delivery riders is falling as networks and systems are optimised with AI. If short-distance truck transportation also becomes unmanned in the future, then another group of jobs will disappear.
The immediate problem is university graduates, because entry-level jobs are increasingly hard to find. This is the case around the world. I don’t think China has reached a point where you can’t find work yet. There will always be work, even if it’s food delivery, but these are not the roles graduates seek. Or appropriate jobs may exist, but they are often not in first-tier cities or the graduates’ hometowns, which are usually the two preferred locations.
Having said that, at present, even with unemployment, basic survival is not threatened for most because general economic conditions have improved. Even if young people go home to “lie flat” and rely on their parents, they won’t go hungry. But I think the biggest and most important problem with AI is the erosion of a sense of purpose. Even if basic needs are met, people will begin to ask whether there is any meaning in this way of living.
My imagination is limited, and I don’t know what new jobs will emerge. New types of demand will certainly appear, but I am sceptical that this will translate into a net increase in employment or successfully lower the unemployment rate on a macro level.
How, then, should governments manage the risks that AI brings to employment?
This brings us back to the same issue of social security. The issue is heavily discussed by academics around the world, but the shock has not yet reached a scale that forces major policy intervention in any country.
In extreme scenarios, people have proposed a universal basic income, but that requires even greater fiscal expenditure. The AI economy cannot sustain itself if workers lose their incomes and can no longer consume.
If everyone lost their jobs overnight, the solution might be clearer – universal basic income. The difficulty lies in the transition: many will lose their livelihoods while a small group becomes exceptionally wealthy. Taxing this wealth and creating alternative employment during this transition period will be difficult for any country. But we are not yet at that point, and you cannot mobilise state resources without an immediate crisis.
And even if we were to reach that point, it will be a completely different world. It’s hard to imagine what that would look like – our relationship with AI, how to tax AI and robots, and so on. These will all change.
But for that to happen, AI must enter the physical world. We need to see functioning, general-purpose robots first, and we are not there yet. While AI has progressed rapidly over the past six months, the marginal rate of development appears to be slowing. We have seen multiple cycles of AI optimism over the past half-century, but only recently did we see a breakthrough. It remains possible that artificial general intelligence or functional robotics will never truly materialise.
More Articles from SCMP
‘Point of grievance’: Philippines urged to rethink school safety approach
Meet Grigor Dimitrov, the Bulgarian tennis player dating 3 Body Problem star Eiza Gonzalez
How snakes evolved to lose their limbs
Chinese teachers publicly shamed on stage for students’ low test scores, sparking outrage
This article originally appeared on the South China Morning Post (www.scmp.com), the leading news media reporting on China and Asia.
Copyright (c) 2026. South China Morning Post Publishers Ltd. All rights reserved.
