核心问题
Which market failure, security constraint, or coordination problem justifies moving resources into this sector rather than another one?
Industrial policy is macro policy when it changes the economy's productive frontier, supply security, and regional income map.
The central question is discipline. A state can coordinate investment, but it must also know when support is failing.
政策目标
Which market failure, security constraint, or coordination problem justifies moving resources into this sector rather than another one?
Industrial policy uses subsidies, procurement, credit, standards, and coordination to build capacity where markets underinvest or strategic risk is high.
Industrial policy succeeds only when the state can identify the bottleneck, discipline recipients, and stop support that is not building capacity.
工具与手段
政策真正可以调动的杠杆。每种工具都有各自的操作目标、信息需求和政治成本。
Subsidies and tax credits
Price support lowers private cost for targeted investment.
Public procurement
Government demand can create a market before private demand is deep enough.
Credit and guarantees
Public finance can unlock projects with high fixed cost or strategic risk.
Standards and coordination
Technical standards, permitting, infrastructure, and training turn isolated firms into an ecosystem.
传导渠道
政策工具影响价格、产出、就业、信贷、对外收支或产能的渠道。
公共资金与私人投资
补贴、税收抵免或政府采购会为某个行业增加需求。问题在于,它是否带来了原本不会发生的私人投资,还是只是为企业本来就会建的工厂买单;若是后者,公共成本是实实在在的,新增产能却很少。
必须持久的支持
工厂、供应商网络和训练有素的劳动力都需要多年才能建成,因此企业只有在预期支持会比建设期更持久时才会投入。2025年7月美国提前终止多项能源税收抵免时,规划长期项目的企业认识到,政策本身就是一种需要计入价格的风险。
国外的反应
对一国生产者的支持会压低其出口价格,并可能招致贸易伙伴的回应,从反补贴税到它们自己的补贴计划。根据世贸组织(WTO)规则,成员可以征收反补贴税,以抵消受补贴进口造成的损害。
时滞
效果究竟何时显现。有些价格会立即调整;数量、合同、招聘和投资通常反应得更晚。
Program design
Months. Eligibility, guardrails, procurement rules, and agency capacity decide the first shape of the policy.
Investment response
Years. Plants, supply chains, workforce pipelines, and learning curves do not move on quarterly clocks.
Productivity payoff
Years to decades. The payoff depends on scale, spillovers, competition, and whether the sector keeps learning.
制度约束
决定一项技术上自洽的政策能否实际使用的法律、政治、资产负债表和操作层面的限制。
State capacity
A subsidy is only as good as the agency that allocates, monitors, and disciplines it.
International rules
Trade rules, subsidy disciplines, and retaliation shape how far the policy can go.
Exit discipline
Without sunset rules or performance tests, temporary support becomes political rent.
失灵方式
政策可能失灵、来得太晚或作用于错误边际的各种方式。
Picking without disciplining
Support can become a transfer to incumbents if performance and exit rules are weak.
Capacity without demand
Building supply does not guarantee a market if standards, buyers, and complementary infrastructure are missing.
Strategic overreach
National-security language can justify too many sectors and dilute focus.
取舍
任何有用的干预都会牵动别的东西。
挑选行业,还是为所有行业改善土壤
对特定行业的支持可以解决某个具体的协调或安全问题,但要求政府知道是哪些行业,并对其余行业说不。基础设施、职业培训和科研经费等普惠措施,对规划者的要求更低,但对结果的控制也更弱。
受益的地方,付钱的纳税人
产业计划会挑选地区和企业,看得见的收益归于建厂的地方。全国各地的纳税人都在付钱,而更广泛的理由则建立在对供应商和其他行业的溢出效应上,这些效应真实存在却难以衡量。
固定条件,还是逐案支持
事先写明的条件,例如终止日期、追回条款以及产量或出口目标,使终止无效的支持成为可能。逐案裁量让主管机构可以根据所学调整做法,但也让受益者更容易游说延续支持。
需要监测的数据
关闭本页之后,继续追踪那些可能推翻这一诊断的数据序列。
- Private investmentBusiness investmentA serious program should crowd in durable investment rather than only replace it.
- ProductivityProductivityCapacity policy should eventually show up in productivity, costs, exports, or resilience.
- Sector employmentManufacturing employmentJobs data show whether capacity is becoming an operating ecosystem.
历史案例
检验这项政策诊断的真实案例。
1960s-1990s
East Asian industrial upgrading
Export discipline, learning, and state coordination remain the reference case for the debate.
2022
Semiconductor capacity race
The CHIPS and Science Act, signed on August 9, 2022, gave the U.S. Department of Commerce $50 billion for semiconductor research, development, and manufacturing, making supply security a macro question.
2022-2025
Green industrial policy
The Inflation Reduction Act, enacted on August 16, 2022, worked through tax credits, among them credits for clean vehicles; a law signed on July 4, 2025 ended those credits for vehicles acquired after September 30, 2025.
The market failure has to be named
Industrial policy is not a license to subsidize whatever sector is politically attractive. The argument has to identify a failure: learning spillovers, coordination failure, missing infrastructure, national security risk, underpriced climate damage, or a capital-market failure for high fixed-cost investment. Research using data on all medium and large Chinese enterprises from 1998 to 2007 found that industrial policies directed at competitive sectors, or designed to foster competition within a sector, increased productivity growth [1].
Naming the failure disciplines the instrument. A coordination failure may call for standards and procurement. A learning spillover may call for temporary support tied to output, exports, or cost reduction. A security constraint may require redundancy rather than ordinary cost minimization. Trade law bounds the choice: the WTO's subsidies agreement divides subsidies into prohibited and actionable ones and lets trading partners offset subsidized imports that injure them [2].
Discipline separates capacity from rent
The hard part is not announcing support. It is monitoring whether support is producing scale, productivity, resilience, or spillovers. Without performance tests, public money can become a transfer to firms that would have invested anyway or to firms that never become competitive. The newer empirical work, which pays closer attention to measurement and causal inference than the correlational studies before it, gives a more positive but more conditional verdict: results depend on context, instruments, and governance [3].
Discipline can mean sunset clauses, clawbacks, competitive grants, export tests, procurement milestones, cost benchmarks, and public reporting. The design should make it possible to stop a program without treating every failure as a political defeat. International rules add some outside reporting: WTO members must notify all specific subsidies to the organization's subsidies committee, with full notifications every three years, and those notifications are reviewed and discussed by the other members [4].
The macro payoff is capacity, not announcements
A subsidy announcement is not macro evidence. The evidence appears later in investment, output, prices, export capacity, domestic supply security, productivity, and local labor-market formation.
The U.S. laws of 2022 are a live test. The CHIPS and Science Act, signed on August 9, 2022, gave the Department of Commerce $50 billion for programs in semiconductor research, development, and manufacturing [5]. Public Law 117-169 of August 16, 2022, known as the Inflation Reduction Act, worked largely through tax credits, among them credits for clean vehicles [6]. Within three years the rules changed again: the budget law signed on July 4, 2025 brought forward the end of several energy credits, so that the clean-vehicle credits do not apply to vehicles acquired after September 30, 2025 [7].
The monitoring dashboard should therefore include both project delivery and macro outcomes. Plants that open late, hire few workers, import most inputs, or fail to lower bottleneck risk are weak evidence for the policy even if the headline spending number is large. And support that can be withdrawn within a few years changes the calculation of every firm deciding whether to build.
情景 · 解读政策的方法
在套用政策框架之前,先核对工具、渠道、时滞和失灵方式
工具
在判断政策立场之前,先说清具体使用的是哪种工具。同一个目标可以通过利率、支出、税收、监管、沟通或资产负债表操作来实现。
传导
追踪政策变动如何经由家庭、企业、银行、市场、预期、汇率和公共部门资产负债表传导。
时滞
区分宣布、实施、市场反应、实体经济反应以及数据发布的时间。
失灵方式
说明哪些情况会让政策适得其反、生效太晚、效果外溢到国外、只是转移风险,或无法触及受约束的部门。