China’s K-shaped economy and its K-shaped Policy Regime

A world-class technology sector sitting on top of a weak economy — and a policy machinery to keep it that way

You have probably seen both halves of China’s economy in the news. One half is the technology story: DeepSeek and Kimi rattling Silicon Valley, Chinese electric-vehicle makers outselling almost everyone, a country that has pushed its way to something close to self-sufficiency in the industries governments everywhere now treat as strategic. The other half is the macro story: a declining property market for years but has yet to bottom out, youth unemployment at record highs, consumers who will not spend, and private investors who hesitate to commit to long-term projects.

China has a K-shaped economy — its upper micro-technology arm pointing up, toward technology, and its lower macro arm pointing down, toward almost everything else that affects the welfare of the ordinary Chinese people. An equivalent framing is that the upper arm is the supply side and its lower arm is the demand side. This will get you to the same place as the analysis above, which is that China has a formidable supply capacity on top of a lackluster domestic final demand.

The official GDP figures reveal but also substantially mask the K-shaped nature of the Chinese economy. Chinese GDP growth rate since 2012 is the weakest ever since China’s economic reforms in 1978 and for the years since 2020, the official GDP releases have been widely questioned by independent analysts. For example, while the official GDP growth rates were reported at 5.2% for 2023 and 5% for 2024, the Rhodium Group puts the real figures closer to 1.5% and 2.8%. A Chinese economist, Gao Shanwen, came up with similar estimates. Debt has climbed to roughly 294% of GDP, nearly triple its pre-2008 level. Real estate, which at its peak was about 30% of the entire economy, is still reeling. Youth unemployment reached 20.4% in 2023, the highest since the series began — after which the number briefly stopped being published; one unofficial estimate put the true figure near 46%. The retail sales have consistently underperformed against overall GDP growth.

The latest numbers are particularly revealing. In the first half of 2026:

• GDP grew 4.7%.

• Manufacturing value added grew 5.6%.

• Exports grew 13.4%.

• Retail sales grew only 1.3%.

• Real per-capita consumer expenditure grew 2.7%.

This indicates that growth has increasingly been driven by production, technology-intensive manufacturing and exports rather than household consumption. China’s statistical agency, the National Bureau of Statistics, itself described the current situation as a conflict between “strong supply and weak demand” (供强需弱).

There can be two dramatically different narratives to explain this K-shaped economy. One is the nature of a market economy. The boon and the bane of a market economy is that it can improve economic performance but do so at the expense of the welfare of the ordinary people. A future dominated by AI is feared to be such an economy. It releases huge productivity gains but it destroys job generation and income growth of citizens. Some have pointed to China’s technological feats as indicative evidence of such an economy in place.

This narrative lacks even surface plausibility. One inevitable implication of this view is that Chinese GDP growth should accelerate, powered by productivity. There is no evidence that this has happened. The headline performance of GDP has slowed, and slowed substantially according to independent estimates. The evidence on productivity is even more damning. The evidence on total factor productivity ranges from marked slowdowns to stagnation or negative. I cannot say that I have surveyed the studies on this topic exhaustively, but to date, I have not seen a single study that shows that TFP has accelerated.

The second explanation is that China’s K-shaped economy is a result of its K-shaped policy regime. To some extent, even a casual observer would arrive at this conclusion without having to do a lot of digging of facts and data, but let me be thorough here and evaluate China’s policy regime using a statistical approach. I first reported my analysis in a report published by Center for Economic Policy Research and I condense and summarize my findings here.

To do that, I ran the Chinese government’s own economic-policy statements through a textual-analysis method called Content Salience Analysis, or CSA, a tool built to measure how much attention a text devotes to a given topic. The idea behind CSA is that salience is not simply whether a topic is mentioned but how it is mentioned. It scores topics against one another on four dimensions: length (how much space the topic is given), specificity (concrete targets and numbers versus vague gestures), placement (whether it leads the document or is buried near the end), and rhetorical force (how emphatic the language around it is). I used multiple large language models to apply CSA technique to China’s policy documents.

The document I scored is the annual plan report of the National Development and Reform Commission — the NDRC, China’s supra-ministerial economic-planning agency — which it presents to the National People’s Congress every March. It is about the closest Chinese equivalent to the American Economic Report of the President: the government’s own yearly account of what it did and what it intends to do. I had the model sort the report’s topics into two categories. The supply side is the production economy: science and technology innovation (科技/技术创新), industrial upgrading and manufacturing, supply-side structural reform, infrastructure, the digital economy and smart manufacturing, exports, energy and raw materials. The demand side is closer to an economy as ordinary people experience it: employment (就业), income and wages (收入), household consumption (消费), social security and healthcare and education, basic livelihood and housing and elderly care, rural development, and — the one I especially want to flag — Hukou reform (户籍改革), the household-registration change that would grant some 600 to 800 million rural Chinese the same entitlements as their urban compatriots. The CSA value ranges from 0 to 10, from a minimal textual salience given to a topic to a maximum textual salience given to a topic.

The policy statement is “Implementation of national economic and social plan last year and draft national economic and social plan for next year,” presented to the annual meeting of the National People’s Congress held every March by the National Development and Reform Commission, NDRC.

Demand and supply-side topics in the NDRC policy statements.

1) Demand-side topics:

就业 (Employment)

收入 (Income and Wage Growth)

消费 (Household Consumption, Retail, Final Demand)

社会保障/医疗/教育 (Social Security, Healthcare, Education)

民生/住房保障/养老 (Basic Livelihood, Housing, Elderly Services)

三农/乡村振兴/脱贫攻坚 (Rural Development, “Three Rurals,” Poverty Alleviation)

户籍改革 (Hukou Reform and Migrant Labor Integration)

2) Supply-side topics:

科技/技术创新 (Science and Technology Innovation)

产业结构升级/制造业发展 (Industrial Upgrading, Manufacturing Transformation)

供给侧结构性改革 (Supply-Side Structural Reforms)

基础设施建设 (Infrastructure and Public Investment)

数字经济/智能制造/高新技术 (Digital Economy, Smart Manufacturing, High-tech Zones)

出口和国际市场 (Exports, Trade Performance)

能源、原材料、石化 (Energy, Raw Materials, Petrochemicals)

Here are the two graphs of the CSA scores that were generated using this textual analytical approach:

Panel (1): Demand- vs supply-side topics

Panel (2): Topics on science and technology vs employment

Notes: LLMs were instructed to group the topics along the lines of production as the proxy of the supply side and income and living standards as the proxy of the demand side and to apply CSA analysis to the texts of the NDRC documents.

The results are stark. Supply-side topics — technology, industrial upgrading, infrastructure —all sit at the top of the NDRC’s attention. Demand-side topics — jobs, income, social security, Hukou reform — sit reliably at the bottom. And the gap is widening. In 2014 the two were almost even: a CSA score of 7.8 for the supply side against 7.5 for the demand side. By 2024 they had pulled apart to 9.5 versus 6.3. The sharpest single contrast is technology against employment. Before 2016, employment outranked science and technology in these reports; since then technology has climbed steadily while employment has slid. Most remarkably in 2023, when youth unemployment spiked to its record high, that development produced no visible bump in the priority the reports give to jobs. Neither the first Trump trade war nor the Covid pandemic moved the trajectory either.

The K-shaped policy regime is remarkably stable: strong on technology, weak on employment and on the Hukou reforms that some economists in China estimate could boost consumption by as much as 30 percent from its current level. (The estimate is probably on the optimistic side and the estimates were produced at a time when Chinese economy was much stronger than it is now.)

In my 2026 book I explained this dynamic in great detail. In my future Substack, I will come back to this question: why China’s underconsumption problem has been long on recognition but short on solutions. The reason is political. A supply-side push is compatible with the power of the state; a demand-side pivot would require ceding some of that power away. Faced with this choice— in the wake of trade wars and Covid — the Chinese government has chosen the top of that K-shaped policy direction every time, again and again. But there is a cost to that policy choice: Every time it does so, it reliably worsens the bottom half of that K-shaped economy, and, by the way, it will also inconveniently degrade the resource base required to fund and subsidize the top half of the Chinese economy.