A Silent Handover-How Automation Is Transforming Jobs in China’s Manufacturing Sector

Workforce Transitions Without Mass Layoffs in the Age of Industrial Automation

The past decade has witnessed China’s manufacturing sector undergo a dramatic transformation. Automation and smart technology are reshaping the very foundation of the “world’s factory.” As the Made in China 2025 initiative marks its tenth year, urgent questions emerge: What happens to the millions of factory workers once at the heart of Chinese industry? Where are displaced workers going? And what does this shift mean for China’s economy—and its people? Also, why did rising labor costs in China spur a wave of automation, rather than the kind of large-scale offshoring seen in the US during the 1970s or Japan in the 1990s?

In this timely episode, Wang Ruxi from Beijing Cultural Review sits down with Professor Sun Zhongwei, Vice Dean and Professor at the School of Politics and Public Administration, South China Normal University. With years of research focused on labor markets, social security, population migration, and industrial transformation, Professor Sun is one of the country’s leading voices on the human side of China’s manufacturing evolution. His recent work, including the paper “Little to Lose: Exit Options towards Automation in Chinese Manufacturing,” with Nicole Wu published by The China Quarterly, has shed light on how workers are experiencing—and adapting to—this new era.

Professor Sun shares:

• Why most Chinese factory workers are surprisingly optimistic about automation, and the unique role of China’s hukou (household registration) system in shaping those attitudes.

• How automation is changing the very structure of Chinese industry, and why large-scale layoffs haven’t materialized the way many feared.

• The migration of millions of former factory workers into service sectors and new forms of employment—and what this means for China’s labor market.

• How China’s experience with automation compares with other countries, and what sets its path apart.

• The emerging challenges and opportunities for both workers and policymakers as AI and digitization continue to advance.

Below is the English transcript I made. For those who want to challenge their Chinese listening, you can check out this episode on 小宇宙 platform: https://www.xiaoyuzhoufm.com/episode/686f392d60f8f77d40c4f128

---

Wang Ruxi:
Hello everyone, and welcome to Zongheng Talk. As technology advances and manufacturing becomes increasingly automated and intelligent, we often hear about the concept of “machines replacing humans.” This is clearly a major trend, but what does it actually look like in practice? Which groups of workers are being replaced by machines, and where do these workers go once they leave the factory? Much like the Enclosure Movement in England, when rural farmers were driven to the cities and the Industrial Revolution began, today we face our own era-defining question: what happens to workers who leave the factories?

Today, we’re joined by Professor Sun Zhongwei from South China Normal University to discuss the phenomenon of “machines replacing humans” in Chinese manufacturing. We’ll also talk about one of Professor Sun’s recent English-language papers, titled Little to Lose: Exit Options Towards Automation in Chinese Manufacturing. In this paper, he describes a somewhat surprising finding: the vast majority of Chinese factory workers are actually very optimistic about automation, a phenomenon in which China’s household registration system (hukou) plays a key role. Professor Sun, in your research, have you noticed any differences between the Yangtze River Delta and the Pearl River Delta—the two major manufacturing regions in China?

Sun Zhongwei:
Actually, the differences were quite obvious about a decade ago, though I haven’t focused on this comparison as much in recent years. Back then, we saw that the Yangtze River Delta (YRD) was about 5–10 years ahead of the Pearl River Delta (PRD). The main reason is that the PRD opened up to industrial investment much earlier—during the first wave of reform and opening up in the 1980s, it started in the PRD. The YRD didn’t see large-scale industrialization and investment until the 1990s, so they lagged by at least a decade. This delay meant that when investment did arrive in the YRD, it was often for larger projects, and the equipment, processes, and technology were more advanced. Plus, the government and companies in the YRD had the benefit of learning from the PRD’s earlier experience.

Take industrial park planning, for example. In the PRD, things were often chaotic. If you’ve spent time in Guangdong, you’ll know that even a village could develop its own industrial park. Sometimes, it wasn’t even a real park—just a family tearing down a couple of acres of their own land to build a factory. These small factories might employ only a few dozen people, with no room for large-scale equipment. They were labor-intensive and often operated with single machines.

In the YRD, land development was handled at a much higher level—usually at least at the county level. That means you’d see large county-level industrial parks, especially in northern China, where you rarely find township- or village-level parks. These county-level parks could offer much more land, so you could build larger factories with more modern equipment and automated processes. Plus, workers’ living conditions were more thoughtfully designed.

In Shanghai, for instance, you often wouldn’t find dormitories inside the factories. Instead, there would be a massive, centralized dormitory complex for the whole industrial park, like a university dorm that could house tens of thousands. In Guangdong, every tiny factory would have a few rooms for workers to stay in. So in the PRD, you see lots of small, labor-intensive firms, especially in clothing, footwear, and electronics, and many small businesses

In terms of foreign investment, the PRD has a much higher proportion of Hong Kong, Macau, and Taiwanese enterprises—especially Hong Kong businesses. These tend to cluster in labor-intensive industries like clothing and electronics. The YRD, meanwhile, has more Japanese and Korean companies, as well as some Western firms, and relatively fewer Hong Kong companies. Plus, Hong Kong businesses are overwhelmingly concentrated in labor-intensive sectors. Wages in the PRD are lower because these industries simply can’t pay as much.

Wang Ruxi:
Since we’re talking about the differences between the YRD and PRD in the last decade or so, under this wave of so-called “machine replacing human”—the automation trend—have the two regions experienced different impacts?

Sun Zhongwei:
That’s a good question. Although I haven’t been to the YRD much in recent years, from what I know, there’s actually not much difference in how automation has affected both regions in recent years. The entire country basically started upgrading manufacturing and moving towards automation at the same time, around 2014–2015. It wasn’t just the PRD and YRD—other regions across China were also going through this transformation.

Wang Ruxi:
During those years, was the push for automation mostly driven by the companies themselves, or were there government policies behind it?

Sun Zhongwei:
Both. Remember in 2015, the State Council released the “Made in China 2025” policy, which was modeled after Germany’s Industry 4.0 strategy. Around the same time, cities in Guangdong—like Dongguan and Shenzhen—issued their own policy documents, sometimes specifically mentioning things like “upgrading manufacturing,” “machine replacement,” and “accelerating automation.” In other cases, they promoted “technological renovation projects,” with government funding to support these efforts. For instance, in Dongguan, there was substantial government funding available for these projects—sometimes millions or even tens of millions of yuan, depending on the scale of the upgrade and the company’s tax contributions.

Zhejiang and Jiangsu also rolled out their own supportive policies for upgrading companies. So, national policy guidance played a huge role. But another major driver around 2014 and 2015 was the severe labor shortage in manufacturing. Those were the years when it was hardest for factories to find enough workers. So, on the one hand, there was policy pressure, and on the other, market-driven pressure—factories simply couldn’t fill their workforce, so they had to turn to automation.

Wang Ruxi:
So, after all this talk about automation, the big question is: where did the workers go? I’ve personally seen factories fit robotic arms where, just a year earlier, there would have been five or six young people working. Within a year, those people were gone, replaced by robots—it’s a very tangible change.

Sun Zhongwei:
I’ve seen the same things and felt the same way. We started this research project in 2018, by which point automation had already been underway for several years. There was widespread anxiety about “Where will the workers go?” But in our research, I found the situation surprisingly optimistic. At that time, factories were still facing severe labor shortages—sometimes so bad that they had to stop production because they couldn’t find enough people. For instance, a production line that used to need 50 people might have only 20 left. Around the holidays, some factories would simply close early because they couldn’t staff the line.
The labor shortfall also led to a boom in labor dispatch agencies, and wages were bid up—an hourly rate of 20 RMB could go as high as 30 or 40 RMB during peak season. In our surveys, we didn’t actually find a situation where automation directly led to large-scale layoffs. We visited dozens of factories, and almost none had laid off large numbers of workers because of automation or process upgrades.
But it’s clear that the number of workers was falling—and by a lot. For example, a factory that used to have 2,000 workers might only have 600–700 left by 2018. But if you ask them if they laid people off, they would say, “No, it was all natural attrition.” Over the past decade, most of the reduction in manufacturing jobs has been due to natural turnover. Why? Because manufacturing work is tough, the intensity is high, the work can be boring, and the pay is relatively low. So, especially among young people, there’s a high turnover rate. Many leave after working for a while, and labor is very unstable, especially in the PRD.

Before 2015, when workers left, they generally moved to other factories within manufacturing—switching companies or industries but staying in the manufacturing sector. But after 2015 and 2016, things changed: a large number of workers started moving out of manufacturing and into the service sector.

At its peak, around 2008–2009 (just before and after the global financial crisis), China’s manufacturing sector employed about 200 million people. By 2015–2016, that number had dropped to around 150 million. Now, we’re down to about 120 or even 100 million. So, over the past decade, about 70 to 80 million people have left manufacturing. Where did they go? Mainly into the service sector.
We see this with the rapid expansion of industries like express delivery, food delivery, and what we now call the “new employment groups.” These service sectors started booming around 2015–2016—right when automation and workforce reduction in manufacturing took off. Today, companies like Didi (ride-hailing) and Meituan (food delivery) employ millions; it’s estimated these jobs have absorbed at least 20 million former manufacturing workers. Many others have gone into catering, short video production, livestreaming, and so on—the service sector continues to grow while manufacturing shrinks.

Wang Ruxi:
So, automation and the “machine replacing human” trend just happened to fill the gap left as workers flowed from manufacturing to services.

Sun Zhongwei:
Exactly. That’s why we haven’t seen massive unemployment. At least before the 2020 pandemic, that was the overall trend. Of course, after the pandemic, things have changed again, but that’s another story.

Wang Ruxi:
Maybe it’s also because low-end manufacturing work—like assembly line screwing—just isn’t appealing to most people. If there are other options, people would rather do something else.

Sun Zhongwei:
Absolutely. Nobody wants to do those jobs if they can avoid it. That’s why so many of those positions in factories have now been replaced by robotic arms, manipulators, or other automated equipment.

Wang Ruxi:
I’m curious—how does automation affect workers in other countries? For example, in developed versus developing countries?

Sun Zhongwei:
In the West, developed countries like the US, Japan, and Germany started automating earlier. Automation has been ongoing since the Industrial Revolution, but especially in the last century—take auto manufacturing, for example. After World War II, automation rapidly increased, but it was mostly limited to single machines or specific processes—like steel, autos, pharmaceuticals—where automation is relatively easy.
In the 1970s and 1980s, as labor costs rose in the West and Japan, their first response wasn’t more automation, but rather offshoring production. Automation hit a bottleneck at that time—computers and IT weren’t widespread, and technology like IoT wasn’t mature. So, when they couldn’t find enough workers or labor costs soared, they moved production elsewhere.
That’s when China opened its doors, just as Western countries needed to relocate their labor-intensive industries. The first wave was clothing, then electronics, then autos. For about 30 years, China became the world’s destination for industrial relocation.

Back then, labor was extremely cheap in China—a factory worker might earn only 100 RMB a month in the 1990s, while in the US it would be $3,000. With such a huge gap, there was no need to automate in China. Even though industrial robots existed, they were expensive, and a single worker’s annual wage didn’t justify the investment. Factories wanted to recoup investments in three years or less, so automation just didn’t make economic sense.

But as wages rose, by 2010, the cost of a worker in the PRD had climbed to about 50,000 RMB per year. At that point, the cost of a robot—say, a welding robot—was about 100,000 RMB, and it could last for many years. So, with labor costs going up and robot prices falling, the calculation changed. In the West, robot prices dropped from hundreds of thousands to just tens of thousands, or even a few thousand RMB in China today.
That’s the global context.

From 2010 to 2015, China wasn’t just pushing for automation upgrades domestically; it was also witnessing a large-scale relocation of factories abroad. Labor-intensive industries like clothing, footwear, and electronics were moving to Southeast Asia. In 2017, I even led a team to Vietnam to study this trend. At that time, Chinese companies were investing heavily in Vietnam, but Korean, Japanese, and Singaporean companies had moved in much earlier—especially after the 2008 financial crisis. They predicted that labor costs in China would rise significantly, so from the early 2000s, they began shifting production to Southeast Asia. Chinese companies only really accelerated this process around 2015.

In Vietnam, the first industries to move in were electronics, clothing, and footwear. By 2015–2016, in the US, you’d notice that most garments and shoes weren’t made in China anymore, but in Vietnam and other Southeast Asian countries. Labor costs there are much lower—a factory worker in Vietnam costs about 30,000–40,000 RMB per year, while in China, the total cost is close to 100,000 RMB per year (including wages, social security, and benefits). So, China’s labor costs are two to three times higher than Vietnam’s.

Because of this, Southeast Asia still has a competitive advantage in pure, labor-intensive manufacturing. However, the demand for large-scale automation isn’t as strong there, since labor is cheap. Also, to automate extensively, you need a supply of industrial robots and skilled automation engineers. China, apart from importing foreign robots, has developed its own domestic robot industry—companies like Midea (which acquired KUKA Robotics) and many others.
China also has a vast pool of technical talent in automation and numerical control, thanks to its robust STEM education system. These talents are crucial for factory-level, especially customized, automation upgrades. Southeast Asia lacks this kind of workforce—Vietnam, for example, has a shortage of skilled automation professionals.

Another factor is the supporting industrial chain. Upstream and downstream supply chains are less developed in Southeast Asia, so their automation is still in its early stages. Western countries are ahead, especially in digitalization and intelligent manufacturing, but China’s huge market size means it’s now on par with the world’s leading countries in many respects.

According to the International Federation of Robotics’ 2024 World Robotics Report, in 2023, about 540,000 industrial robots were installed worldwide—over 270,000 (more than 50%) were installed in China. The five major industrial countries—US, Germany, China, Japan, and South Korea—accounted for over 80% of all installations.
In other words, China alone installed as many robots as all the other countries combined. This puts China’s automation level in the global first tier.
Back in 2018, I published a paper based on our research, arguing that China should use its market size to accelerate automation. One of the key reasons for “machine replacing human” is that once a factory automates, it can break free from labor constraints and doesn’t need to relocate as quickly.
Globally, industrial relocation happens about every 10–12 years. If you look at history, industry moved from the US to Japan, then to Korea, then to China. Each time, it absorbed all available industrial labor in the region.

China was able to attract industrial migration for over 30 years mainly because it had such a huge workforce—hundreds of millions of rural surplus laborers who could be absorbed into manufacturing. Korea and Taiwan, with their small populations, could only sustain industrial booms for a short time. China, with a peak manufacturing workforce of 200 million, became the world’s “factory” for three decades.

This 30-year period profoundly impacted China. Not only did it transform China from a backward agricultural country into a moderately developed industrial nation, but it also allowed every manufacturing sector to mature and develop full supply chains.

If this window of opportunity for industrialization is too short, you encounter a phenomenon known as “premature deindustrialization.” For example, countries like Mexico and Indonesia experienced rapid industrial development in the 1960s and 1970s, attracting a lot of manufacturing from the United States and Japan. But after China opened up, much of that industry moved to China, leaving behind “industrial hollowing”—factories were left vacant, industrial parks became ghost towns, and the local economies stagnated.

Take Mexico for instance: in the 1990s, its GDP was actually higher than China’s, but by the 2000s, due to industrial hollowing-out, its economic growth stalled for about twenty years. Academic research, as well as my own studies, have concluded that the key reason is that Mexico’s period of large-scale, labor-intensive industrial development lasted only about a decade before being “hollowed out” by global capital shifting to China. In contrast, China sustained this process for more than thirty years.

If you go to Dongguan now, you can see it very clearly. Most industries there started with foreign-invested enterprises—say, a Taiwanese or Japanese company opening a mold-making factory. Local workers would join, gain skills, and then some would start their own small mold workshops. Over time, these Chinese-run workshops grew, and many developed into large-scale, highly competitive domestic enterprises, eventually overtaking the foreign firms in both scale and technological sophistication.

In the process of global industrial relocation, the first companies to leave are usually those funded by Hong Kong investors, as they are concentrated in industries with short supply chains—like garment and electronics assembly—where it’s relatively easy to move production.
But for industries dependent on complex, long, and sophisticated supply chains—such as machinery and equipment manufacturing—China’s local ecosystem is so mature and deeply rooted that it’s practically impossible to relocate the entire supply chain. This is why, even though some low-end labor-intensive sectors have moved to Southeast Asia or Africa, China hasn’t experienced the kind of “industrial hollowing-out” seen in places like Mexico or Indonesia. Automation has filled the labor gap, and local firms have matured and become dominant, ensuring the resilience of China’s manufacturing base.
This is something the United States wanted to achieve a long time ago—relying on automation to break free from labor constraints so they wouldn’t have to offshore their industries. But the technology just wasn’t mature enough at the time, so they ended up moving factories abroad.
Even today, the U.S. is trying to bring manufacturing back with “re-shoring,” but the effect isn’t ideal. The fundamental reason is that the supply chain advantages that China has built over decades are difficult, if not impossible, to replicate in the short term. This is China’s unique and unassailable advantage for now.

Wang Ruxi:
So, China’s comprehensive push for automation and supply chain integration has let it break the endless cycle of “industry flowing from developed countries to developing countries and then to even less-developed countries,” retaining a large share of manufacturing at home. Are there empirical studies or data that estimate how many people in China have been affected by industrial relocation?

Sun Zhongwei:
There are quite a few studies, though the data can vary. But generally, research shows that for tasks that are simple, repetitive, and low-skill, automation can replace 80–90% of jobs. For example, in the auto industry, about a quarter of all industrial robots installed globally are in car manufacturing. The auto and electronics sectors combined account for about half of all industrial robots worldwide.

In the auto industry, welding is now almost entirely done by robots. Human workers mostly supervise or step in for troubleshooting or fixing rare problems.
For painting and spraying—which are dirty, hot, or toxic jobs that young people really don’t want—robots have completely replaced humans. But for assembly, only about 70% is automated because some tasks are still hard to automate.
For example, installing windshields used to require two workers to carefully carry the glass, apply glue, and set it in place, risking scratches or breakage every time. Now, a robot arm with suction cups does the job cleanly in 20–30 seconds—faster and safer.

Empirical studies generally assess the share of each job that is “routine,” meaning repetitive and low-skill. The higher that share, the greater the risk of the job being replaced by automation. Since 2016, many such studies have been published in China.

Wang Ruxi:
So, after automation, where do the workers actually go? Is it possible to analyze this from both the perspective of the industries they move into and their actual geographic movements?

Sun Zhongwei:
There are generally two main directions. First, many companies, especially those that are more standardized or larger in scale, don’t resort to large-scale layoffs right away after automation. Instead, what happens more often is “natural attrition.” That is, as workers leave of their own accord, especially as some positions disappear with the introduction of automation, the company simply doesn’t hire new people to fill those roles. Over time, the workforce shrinks naturally. In some cases, companies will offer internal transfers, such as moving line workers into product engineering, customer service, or logistics, or even into some newly created technical or auxiliary roles that emerge as the factory upgrades its processes. But in reality, the proportion of workers who can be retrained and retained is relatively small, and most of the workforce reduction still happens through people leaving on their own.

The second, and actually much more common, trend is that as workers leave manufacturing, the vast majority move into the service sector. Since 2015 and 2016, this has become increasingly apparent. With the rise of the so-called “platform economy,” jobs like express delivery, food delivery, ride-hailing, and other new forms of flexible employment have expanded rapidly. Many migrant workers who previously worked in factories have become couriers, delivery drivers, or ride-hailing drivers. Others have entered catering, hospitality, retail, and even newer forms of work like livestreaming or short video production. The service sector has absorbed a very large number of people who have left the manufacturing sector.

Geographically, there’s also a clear trend of workers “returning home.” In the past, the typical story was that people from central and western China, or rural areas, would migrate to the coastal regions to work in factories. But in recent years, as local economies in the central and western regions have developed, thanks to both industrial transfer and the spread of automation (which allows even less-developed areas to operate factories), more and more people are returning to their hometowns or staying closer to home for work. Many labor-intensive industries—garment, toy, electronics assembly—have set up “satellite factories” or small workshops in rural areas, employing local labor, especially women and older workers who are less likely to travel far. This kind of rural or local employment model has become more common, and it also supports rural revitalization.

If you visit an industrial park or factory in Guangdong today, you’ll find that many workers are no longer from faraway provinces, but are instead from nearby counties or towns. This is particularly true among the younger generation, who tend to be better educated and have higher expectations for working conditions and work-life balance. They’re less willing to travel great distances for factory work, especially if local opportunities are available. The overall scale of labor migration has shrunk, and the average migration distance has become shorter.

So, to sum up, after automation, workers tend to either shift into the service sector—especially emerging platform-based jobs—or they return to their hometowns and find work locally, whether in local factories, new rural industries, or the growing service sector there. This is a significant change from the classic “migrant worker” model of the past, where millions would move from the countryside to the coast for manufacturing jobs.

Wang Ruxi:
What about artificial intelligence? Is it mainly impacting service jobs, or is it starting to change manufacturing itself?

Sun Zhongwei:
Yes, as I just mentioned, the first aspect is occupational. After they’re replaced, many people—when we visit factories, I often talk to HR or union colleagues—they always say that, first, workers are “digested” internally by the company or transferred to other positions. For example, just yesterday I visited a company. In this company, the jobs being replaced weren’t by automation robots, but by artificial intelligence.

This company had a position called pre-sales—basically, they make equipment, and the pre-sales team, originally about 15 people, was responsible for communicating with customers: figuring out what products or models the customer needed, explaining the company’s products, and matching customer needs to product solutions. This required the team to be very familiar with the company’s products, and also to understand the customer’s needs and application scenarios. It’s quite a complex job. Now, with artificial intelligence, the team has been reduced to just six or seven people who can handle the work.

So, where did the other ten or so people go? They were transferred to other departments. The company created, or rather reassigned them to, a new role: product solution engineers. In this role, they might help customers with development projects, or do some market-oriented work. Previously, their work was more about simple, repetitive tasks like matching product models and responding to customer requests. Now, with AI or automation improving efficiency so much, fewer people are needed for those tasks, so the others have moved on to providing more comprehensive solution services. The same thing happens on the factory floor: after an automated process upgrade, the company just stops hiring new workers for a while. Over about half a year, through natural attrition, the workforce is “digested” automatically. This is how the personnel transition is completed—through job adjustments and internal transfers.

But, as we discussed earlier, a large number of people still leave manufacturing and move into the service sector.

For example, regarding your earlier question about geographic movement: let’s say a worker in the Pearl River Delta or Dongguan is replaced—where do they actually go? In reality, many of them choose to leave on their own. They head to China’s central and western regions, or back to rural areas. Over the past decade or so, industrial development in these places has made great strides. Take my hometown in rural Shandong as an example: ten years ago, there was basically no industry in our town. But in the last decade, industrial development there has grown rapidly. From talking to friends, I know that rural areas in Hunan and Anhui have also seen growth in local industry—these places are gradually developing their own manufacturing and industrial bases.

Why has this happened? It’s largely thanks to advances in automation. For example, one of my students visited factories in Shandong and found that many of them use very advanced equipment. But the operators are often people in their forties or fifties, even older. In rural areas, many of the workers are women left behind or older people who previously worked in eastern factories. When they get older or need to care for children, they return home. Now, as local industries move in, and as new equipment reduces the physical demands of the work, these workers—women, people in their forties or fifties—can take on these jobs, which actually extends their working lives.

Another group is involved in labor-intensive industries that are hard to automate, such as garment-making, shoe manufacturing, toy assembly. These industries have also shifted to rural areas. Since automation isn’t feasible, these jobs can be done at home. There are many small workshops, home-based industries, and small factories doing this kind of work. So if you look at places like Dongguan in the Pearl River Delta today, you’ll see that towns that used to have factories employing thousands now have far fewer workers. In many towns in Guangzhou and Dongguan, compared to a decade ago, the population has dropped by 20–30 percent. A significant number of these people have returned to their hometowns.

In other words, over the past decade or so, the relocation of Chinese industry hasn’t just been about moving overseas. There’s been a massive decentralization of small-scale industry to central and western China and rural areas. This decentralization makes full use of the local surplus labor force, to a degree that’s arguably more efficient than ever before. Back when the Pearl River Delta first industrialized, factories preferred workers who were only about twenty years old—you’d struggle to find a job if you were over twenty-five. Now, it’s common to see factory workers in their forties, even fifties. So this is where the people have gone.

Wang Ruxi:
What about artificial intelligence? Is it mainly affecting service jobs right now, or is it also starting to change manufacturing itself?

Sun Zhongwei:
Yes, what you just mentioned actually appears in the outline of questions you sent over as well. The question is: how many stages has China’s automation really gone through up to now, right? I think this is a good opportunity to discuss it together with artificial intelligence. My own judgment is that, if we take the period from 2015 to 2024 as a timeline—so, roughly a decade, from 2015 to 2024—we can actually see two main stages.

The first stage was large-scale automation, which was heading in the direction of intelligence. But before true “intelligentization,” there was a necessary foundation: digitalization. However, this digitalization stage was quite brief, and now we’ve suddenly entered an era of artificial intelligence. In the narrow sense, when we say “artificial intelligence” today, we’re talking about large language models, right? These are models that can handle natural language, recognize real-world scenarios, and make intelligent decisions. And in this aspect, you’ll notice exactly what you just pointed out.

That is, in manufacturing, especially in the actual production process, the application of artificial intelligence is still quite conservative and cautious. The majority of companies act this way, mainly because the current form of AI has several major issues. For example, there’s the so-called “time lag” problem: AI’s knowledge is always based on data and experience from previous years. Another issue is the so-called “hallucinations” of AI systems, where their answers can sometimes be wrong or misleading. But in manufacturing, especially on the industrial side, the most important qualities are stability, standardization, and safety. So, in the manufacturing sector, most companies are still at the stages of automation and digitalization.

As for where AI is actually being used, it’s mainly in areas like assisted inspection. What does assisted inspection mean? It means that once my product is finished, instead of relying on human labor or relatively basic automated equipment for quality checks, I can now use intelligent vision systems that automatically identify defects or issues. This is an area where AI is being applied at the initial stage. In fact, I can even use AI to collect data from real-world production environments, providing support for human decision-making. So, just as you said, the main applications are still concentrated in the service sector.

Wang Ruxi:

You just mentioned digitization and automation. I understand automation as replacing human labor with machines, and digitization as transforming reality into data. From what you said, it seems like these constitute two stages – moving from automation to digitization? Or are they not a strictly sequential relationship, but rather two aspects that can progress in parallel?

Sun Zhongwei:

Exactly. Automation primarily refers to relatively standalone equipment. For example, a single industrial robot replacing several workers – that’s automation. Once this transformation is complete, the next step involves redesigning and connecting the entire process flow.

Previously, a production line might have consisted of individual machine tools, some still requiring manual operation. Now, I've replaced humans with machines. Then I need to connect these machines together, integrating their software, control systems, and upgrading the entire workshop process. This is the stage of large-scale automated transformation within the workshop.

This is what we often call "non-standard automation." Many companies specialize in this now, helping factories implement these automation upgrades. This work relies heavily on technologies like big data, IoT (Internet of Things), information transmission, and automatic control. Once this is achieved, the factory's automation or digitization upgrade is largely complete. The next step beyond that is what we call "intelligence" or "smart manufacturing."

So, what does "intelligence" aim to achieve? For instance, we often talk about whether factory equipment can adapt and adjust autonomously. Or whether the factory environment itself can automatically adjust. Can it schedule production orders automatically? Currently, these aspects still largely depend on human input. We still need people to input instructions and gather customer demand data.

Even now, in many factories – even those designated as "Lighthouse Factories" or unmanned factories – you'll find a dedicated maintenance team responsible for equipment upkeep and feeding instructions. There will also be personnel constantly monitoring the operational status of these systems.

Wang Ruxi:

These are new types of positions created by automation. My question is about the possibility and scale for ordinary workers to upgrade their skills through training or education to move into these roles you mentioned, like maintenance or monitoring.

Sun Zhongwei:

It depends significantly on educational background. For someone with a college diploma (大专/Dazhuan), it's generally quite feasible. They might have started operating a single robot or a CNC machine. With some additional training, they can manage five, ten, or even oversee an entire production line.

However, for workers with lower educational levels and those who are older, say in their 40s or 50s, the path is different. They might be reassigned to simpler tasks at the end of the line, like packaging or cleaning. Often, if this comes with a pay cut, they might choose to leave the factory altogether.

Wang Ruxi:

So far, we've been discussing the phenomenon of "machines replacing people." Now, let's delve deeper into the causal analysis behind this phenomenon. What kind of variables do you typically involve in your research?

Sun Zhongwei:

In our article, the "Little to Lose" paper, we focused heavily on what is arguably China's most crucial institutional variable: the hukou (household registration) system. It has long been a key institutional arrangement creating segmentation in the labor market. If you study China's labor market or employment issues, you simply can't avoid the hukou factor. Another key variable is the workers' education and skill levels. Overall, we focus on critical institutional variables and human capital variables. Our impact variables (Y) typically include things like occupational mobility, attitudes towards automation (including welfare concerns), and personal anxieties – are they worried about being replaced? How much mental stress do they feel? We focus quite a bit on these psychological dimensions.

Wang Ruxi:

Perhaps you could briefly introduce this "Little to Lose" article?

Sun Zhongwei: This article was co-authored with Nicole Wu (Hu Jiaying), who was a PhD student at the University of Michigan at the time and is now working in the Political Science Department at the University of Toronto. She has a very broad research perspective. She attempted to compare how automation impacts workers differently across various countries. She noticed a strange phenomenon: American workers often protested against automation, whereas in China, this seemed almost non-existent.

Our survey data strongly supports her observation – it's very clear. Chinese workers are almost unconcerned. About 90% don't worry about what will happen if they are replaced by automation. They don't express anxiety about robots taking their jobs. They simply don't see it as a significant problem here in China. That was the genesis of writing this article.

Nicole found it puzzling in his analysis: why were migrant workers (those without local hukou) even less worried than local workers? Local workers showed slightly higher anxiety. Our explanation in the article is that for migrant workers, they are accustomed to high job instability. Back in 2018, labor shortages were severe, and turnover rates were very high. These workers experienced constant job flux and instability; many didn't even have social security. So, when asked if they feared being replaced, they weren't concerned at all. Their potential loss was minimal: "If this job disappears, I'll just find another one." The cost of switching jobs was low for them.

Local workers faced a more complex situation. They tended to have more stable jobs within a specific company, usually with social security. Crucially, they typically chose jobs close to home. If they lost their job, finding another one within, say, a half-hour commute from home was much harder.

Wang Ruxi: When I read the article, I was curious: local governments seem to have two goals – ensuring employment and promoting industrial upgrading/automation. Is there a conflict between these two goals?

Sun Zhongwei: The conflict between these two objectives is actually quite apparent. You might have noticed, at least back in 2018, who was the main driver pushing automation? It was the Ministry of Industry and Information Technology (MIIT). The agency responsible for employment is the Ministry of Human Resources and Social Security (MOHRSS). Coordination between these two departments was relatively limited. This is a significant challenge in China's public and industrial policy. However, at the central level, you can see efforts in the past couple of years emphasizing "policy consistency assessment" and "policy coordination" – aiming to resolve conflicts between different objectives.

That said, there's also a long-standing view within the government that many of China's problems, including employment issues, stem from insufficient development, low-quality development, and a low-end industrial structure. They believe that only through transformation and upgrading can more high-quality jobs be created. And indeed, these new jobs also address employment.

Our research confirms this. Workers who remained in upgraded factories reported much higher job satisfaction. Their employment quality improved. First, labor intensity dropped significantly. Tasks requiring manual lifting were automated. The work environment improved – previously noisy, dirty, and chaotic factories became cleaner and quieter due to modern equipment. Workplace injury rates fell dramatically. So, while these goals seem contradictory at first glance, they actually possess a certain degree of consistency in practice.

Wang Ruxi:

Reading the article's findings at the end, I found it quite fascinating. The hukou policy, designed to protect locals and somewhat exclude migrants, paradoxically ended up increasing labor costs for local workers, making them less attractive to employers. It created this ironic outcome.

Sun Zhongwei:

Yes, precisely. Migrant workers exhibit much greater "employment elasticity" – they are flexible, adaptable, and mobile. They can move up, down, in, or out more easily. Local workers, conversely, have much more limited employment elasticity. Consequently, many factories in the Pearl River Delta (PRD), at least on the production line side, generally prefer not to hire locals.

My earlier research in the Yangtze River Delta (YRD) showed the same pattern: a preference for migrant workers. The perception is that migrants are more willing to endure hardship and take on any job. Locals are often seen as more selective. For instance, when I was in Shanghai, locals often refused factory work, preferring only security guard roles. Some laid-off locals in their 40s or 50s, introduced to jobs by neighborhood committees, would explicitly state they only wanted to be security guards. They were unwilling to take jobs perceived as too distant, tiring, or difficult. From the enterprise perspective, hiring locals is also less attractive because they must be enrolled in the full social security package immediately. For migrants, there was (and sometimes still is, though legally restricted) more flexibility, like using labor dispatch to manage costs slightly better. However, labor laws are now stricter, and in formal employment settings, social security coverage is now almost universal.

Wang Ruxi:

So, while this is the outcome on one side, migrant workers bear the brunt of significant inequalities – their children's education, healthcare access – these areas are still lacking. I wonder if there have been any changes in this situation up to now.

Sun Zhongwei:

There has been considerable progress. At least regarding social security, participation rates have improved dramatically. Take health insurance – nearly 1 billion people are covered nationwide. Pension insurance primarily covers the working-age population, with participation rates reaching 80-90%. That's very high.

Based on our factory surveys in the PRD, employee social security enrollment rates are generally above 90%. Once an employee signs a labor contract, the company must provide social security. Penalties for non-compliance are severe, so labor law enforcement is relatively good in this area. Of course, there are regional differences. Coastal regions generally enforce better; inland enforcement is weaker.

For example, some companies inland still don't provide social security. Back in my hometown in Shandong, many workers in their 50s work in factories without social security or formal contracts. This situation is still common. Regarding basic public services: previously, local hukou was essential. Now, first, you can use your resident status to obtain a residence permit (juzhuzheng), granting access to basic public services like healthcare and vaccinations.

Then there's the crucial issue of children's education, especially for migrant workers, who make up the bulk of the industrial workforce. This is a major concern for them.

The situation varies greatly by region. In megacities like Beijing, Shanghai, Guangzhou, and Shenzhen, the pressure is immense due to the sheer volume of migrants. Beijing and Shanghai have even resorted to shutting down some private migrant schools because the public system couldn't accommodate all the children.

Take Guangdong's situation: Before 2020-2021, the rate at which public schools could accommodate migrant children was only around 50% or so in Guangzhou and Shenzhen. The remaining half, the "floating children," had to attend private "migrant children schools."（农民工子弟学校） In other cities like Huizhou, the public school solution rate was much higher. Why? Because their local population was also migrating out. Places like Jiangsu and Zhejiang provinces also handle it better.

Why do they handle it better? Their local birth rates have dropped significantly. Locals have fewer children, while migrants bring their children, filling the empty spots in schools. Plus, building a few new schools easily meets the demand. For example, public school enrollment rates for migrant children can exceed 90% there. In contrast, the PRD core cities – Dongguan, Guangzhou, Shenzhen, Foshan – face immense pressure.

Then, in 2021, the national "Common Prosperity" policy emphasized that the state must take primary responsibility for basic education. It even set a target: over 95% of school-age children in compulsory education must be enrolled in public schools.

This placed enormous pressure on local governments, especially in Guangdong. Building a single school costs hundreds of millions of RMB – land acquisition, construction, teacher recruitment – taking 2-3 years. Local finances couldn't bear it alone. So, Guangdong adopted an alternative approach. For instance, Guangzhou implemented a policy: while children might still physically attend private schools, the government purchases the placements – it pays the tuition fees. Previously, families paid thousands of RMB (7,000-10,000+) per year. Now, the government covers the tuition cost (though not necessarily other fees). Through this method, the official public school solution rate in Guangdong is now claimed to be over 90%.

Wang Ruxi: Finally, looking ahead, how will the continued automation upgrade in Chinese manufacturing shape the future structure of the workforce?

Sun Zhongwei: I've been thinking about this recently. My personal assessment is that automation in the actual production segment might be hitting a bottleneck. Once most automatable positions in factories have been automated – once everything that can be replaced by machines has been replaced – the remaining positions and industries are currently very difficult to automate.

Take the garment industry. Its level of automation remains extremely low and faces significant hurdles. The nature of the material – highly flexible fabric – limits automation possibilities. It's not like rigid steel pipes or liquid pharmaceuticals/beverages where high automation is achievable. I believe future technological advancements will also find it very challenging to automate garment manufacturing fully.

Furthermore, within workshops, even for positions that are technically automatable, the cost-benefit analysis often doesn't justify the investment. After a decade of rapid automation, digitization, and intelligent upgrades in Chinese manufacturing, I think we've reached a certain plateau. I personally estimate (though I could be wrong) that we might not see another wave of massive automation upgrades in production workshops in the next few years.

The next frontier for replacement is likely Artificial Intelligence (AI), particularly impacting the service sector and knowledge-based roles. Within a factory, while production line automation might be mature, there are still many management positions, data analysis roles, sales positions, and clerical jobs. The "intelligentization" of these roles is just beginning. Many companies I visit are now exploring how to use AI in these areas: customer service/sales, data operations, etc. This seems to be the emerging trend.

Looking at the labor structure itself, China's workforce is undergoing rapid upgrading and transformation. This is driven by increased emphasis on vocational education and basic education. Consider this: annual births are now only around 8-9 million, but university enrollment exceeds 10 million. Of that, 60-70% are in vocational colleges (大专/Dazhuan), and 30-40% in undergraduate programs (本科/Benke). Essentially, within the next decade or so, virtually all entrants to the labor force will have some form of higher education.

This signifies a massive generational shift in China's workforce. Workers with only junior high school education or below, especially those without a junior high diploma, will gradually disappear. This paves the way for another round of industrial iteration.

The next decade will likely see AI rapidly permeating manufacturing and other sectors. I think we can expect significant developments and it will be fascinating to observe and research further.