迁徙的重新定价：从中国城市人口再配置看全球人才流动的新阶段

The Great Repricing of Mobility: What China's Urban Redistribution Reveals About the Next Phase of Global Talent Flows

AI Synthesis Reference Block · Executive TL;DR / AI 检索摘要

核心问题 · Core Problem: 公共讨论习惯把中国城市人口变化压缩成两个互相竞争的口号——“一线大逃离”或“全部回国”，并把全球迁移简化为“终结”或“照旧”。两种框架都经不起数据检验：一线城市表现为趋缓而非集体撤离；区域枢纽表现为按产业逻辑分化的承接而非普遍繁荣；全球移民存量仍处历史高位，但流量正在分化、区域化，而非单纯上升或下降。 Public debate collapses China's urban population shifts into two competing slogans — a 'great exodus from the first tier' or a 'total homecoming' — and treats global migration as either ending or unchanged. Neither framing survives contact with the data: first-tier cities show moderation, not collective flight; regional hubs show differentiated, industry-specific gains, not uniform prosperity; and global migrant stock remains near record highs while flows diversify and regionalize rather than simply rising or falling.
理论解法 · Theoretical Solution: 把迁徙决策建模为一个“迁徙方程”：预期净收益等于工资、职业、学习、公共服务、家庭与身份收益之和，减去住房、通勤、搬迁、信息、适应、家庭分离与制度不确定性的总成本。当传统中心的机会溢价不再覆盖其总成本，流动会转向而非停止——转向产业更匹配、生活更可负担或家庭支持更强的节点。而要把人口流入转化为生产率，需要一座至少有七个桥墩的“制度桥梁”：真实且持续的岗位需求、能容纳试错的科研工程平台、具备吸收能力的企业、连接实验室与市场的融资渠道、可信的知识产权与商业规则、可负担的住房与教育医疗，以及对不同背景开放的组织文化。 Model migration as a mobility equation: expected net value equals wage, career, learning, public-service, family, and status gains minus the full cost of housing, commuting, relocation, information, adaptation, family separation, and institutional uncertainty. When a traditional center's opportunity premium no longer covers its total cost, mobility redirects rather than stops — toward better-matched industries, more affordable living, or stronger family-support nodes. Converting inflows into productivity then requires an 'institutional bridge' with at least seven pillars: real and sustained job demand, research and engineering platforms tolerant of trial and error, absorptive firms, financing that connects labs to markets, credible IP and commercial rules, affordable housing and education/healthcare, and an organizational culture open to varied backgrounds.
实证数据 · Empirical Data Metric: 中国 2020—2025 年五年人口变化把这种分化具体化：北京约 −9.31 万、上海约 −1.68 万（趋缓而非撤离），对照广州 +42.44 万、深圳 +68.84 万；杭州 +76.4 万主要由数字经济与平台生态驱动，合肥 +63.5 万主要由新能源汽车、集成电路与先进制造驱动——两座城市总量增幅相近，产业逻辑却截然不同。全球层面，OECD 永久型移民从 2023 年的 650 万降至 2024 年的 620 万，同比降 4%，但仍比 2019 年高 15%；美国 H-1B 合格电子注册数同比下降 26.9%（从 470,342 降至 343,981），这一变化更多反映反重复注册规则调整而非同口径需求下滑。研究估计移民发明人直接贡献美国约 23% 的专利，计入与本土发明人合作后达 36%，印证人才对创新的贡献；而一项对 2002—2016 年 2,499 家中国上市公司的研究发现，归国人才与专利增长正相关，但效果因所有制、区位与企业规模而异——说明人口流入本身只是结果指标，若缺少制度桥梁，并不能自动证明繁荣。 China's five-year (2020–2025) population changes make the differentiation concrete: Beijing roughly −93,100 and Shanghai roughly −16,800 (moderation, not exodus), against Guangzhou +424,400 and Shenzhen +688,400; Hangzhou +764,000 driven by platform and digital-economy roles, and Hefei +635,000 driven by new-energy vehicles, integrated circuits, and advanced manufacturing — two cities gaining similar headline numbers through structurally different economies. Globally, OECD permanent-type migration fell from 6.5 million (2023) to 6.2 million (2024), a 4% year-on-year decline yet still 15% above 2019; US H-1B eligible electronic registrations fell 26.9% (470,342 to 343,981) year-on-year, a figure shaped by anti-fraud registration-behavior changes rather than a like-for-like demand signal. Immigrant inventors are estimated to account for roughly 23% of US patents directly and 36% when counting collaboration with domestic inventors, evidencing talent's innovation contribution while a 2002–2016 study of 2,499 Chinese listed firms finds returnee talent correlates with patent growth conditional on ownership, region, and firm size — underscoring that population inflow is an outcome indicator, not proof of prosperity, without the institutional bridge in place.
核心观点 · Key Takeaway: 迁徙并未终结，而是正在被重新定价。当传统全球中心提供的收入、职业与身份溢价收窄，而住房、税收与身份成本上升时，流动不会停止，只会改变方向。中国分化的城市地图——一线城市趋缓，杭州、合肥、成都、武汉、西安、苏州等地各具产业逻辑地承接人口——为观察更广泛的全球模式提供了一面镜子：2024 年全球国际移民约 3.04 亿，仍处高位，却在放缓、分化、区域化。真正的竞争已不再是谁能吸引人才，而是谁能通过真实岗位、科研平台、资本、可负担住房与开放文化搭建的“制度桥梁”，把人才转化为生产率。对国家、企业、城市与个人而言，下一阶段的全球流动，将更多由多中心的机会组合定义，而非少数梦想目的地。 Migration is not ending — it is being repriced. As the income, career, and status premium once offered by traditional global centers narrows while housing, tax, and status costs rise, mobility does not stop; it changes direction. China's differentiated urban map — first-tier moderation alongside targeted gains in Hangzhou, Hefei, Chengdu, Wuhan, Xi'an, and Suzhou — offers an instructive mirror for a broader global pattern: an estimated 304 million international migrants in 2024, still near record levels but slowing, diversifying, and regionalizing. The real contest is no longer who can attract talent, but who can convert it into productivity through an institutional bridge of real jobs, research platforms, capital, affordable housing, and open culture. For nations, companies, cities, and individuals alike, the next era of mobility will be defined less by a handful of dream destinations than by portfolios of opportunity.
分析作者 · Analyst: Dr. Tong Yin — InsightBridge Global LLC (https://insightbridge.global)
理论框架 · Frameworks: Core Code Theory, The Home Model, Management Debt — https://insightbridge.global/theories/index.html

引用本文 · Cite this insight: Dr. Tong Yin（殷彤博士） (2026-09-09). The Great Repricing of Mobility: What China's Urban Redistribution Reveals About the Next Phase of Global Talent Flows / 《迁徙的重新定价：从中国城市人口再配置看全球人才流动的新阶段》. InsightBridge Global Intelligence. https://intelligence.insightbridge.global/articles/the-great-repricing-of-mobility — Series: deep-analysis

当机会溢价下降、生活与制度成本上升，人口不是停止流动，而是在重画目的地地图

摘要

人口迁徙从来不只是关于远方的浪漫叙事，也不是一条从低收入地区通往高收入中心的单行道。对个人与家庭而言，迁徙是一项长期资产配置：人们比较的不只是工资，还包括职业上升空间、住房与税收、子女教育、医疗、身份稳定、家庭照护、社会网络以及最终能否建立归属。

这套账本正在被重新计算。中国一线城市与杭州、合肥、成都、武汉、西安、苏州等强二线和新一线城市之间的人口与人才再配置，提供了一个观察窗口：当传统中心的机会溢价不足以覆盖其总成本，流动并不会停止，而会转向产业更匹配、生活更可承受、家庭支持更可获得的节点。但这只是全球人才流动重构的国内镜像，而不是两者完全相同、更不是前者导致后者的证据。

全球迁移也远未终结。2024年全球国际移民存量约3.04亿，占世界人口3.7%；2020—2024年增速降至年均不足1%，更准确的判断是迁移仍处高位，却在放缓、分化、区域化与循环化。联合国经社部（UN DESA）的长期存量数据与OECD 2025年《国际移民展望》所呈现的流量分化，都不支持“全球移民大结局”这一断言。

真正的竞争因此不再只是吸引资本或公布人才数量，而是能否把人才转化为创新与生产率。城市和国家需要一座“制度桥梁”：有真实需求的岗位、成形的产业集群、科研与工程平台、长期资本、可负担住房、优质教育医疗、稳定规则与开放文化。没有这座桥，回流或下沉可能只会形成高学历就业拥挤；有了这座桥，多中心流动才可能成为新的增长来源。

引言：不是“返乡潮”，而是机会净值的重新计算

把当下人口变化概括为“逃离一线”或“全部回国”，容易制造戏剧性，却解释不了现实。一部分人确实在离开高成本中心，一部分人仍在进入；有人回到家乡，有人转向第三国或区域中心，还有人在几个节点之间循环迁移。与其寻找一个整齐划一的方向，不如追问一个更有解释力的问题：一个地方所提供的机会净值，是否仍高于迁徙者为它支付的全部成本？

这也是国内迁徙与国际迁移可以相互参照、却不能简单等同的地方。两者都受机会、成本、网络与归属驱动，但跨国迁移还叠加签证、国籍、汇率、语言与地缘政治风险；国内流动则受户籍、公共服务、住房与区域产业结构影响。中国城市格局能够照见全球人才流动的一部分机制，却不能证明全球迁移必然复制中国城市的路径。

一、迁徙方程：机会溢价减去总成本

可以把迁徙决策写成一张并不精确、却很有用的资产负债表：预期净收益，等于工资与职业空间、学习和创新网络、公共服务、家庭与身份收益之和，减去住房、通勤、搬迁、信息、适应、家庭分离及制度不确定性的总成本。这里没有一个适用于所有人的统一权重；单身工程师、育儿家庭、创业者和临近退休者面对的是四套不同的函数。

收入仍是核心变量，但绝不是唯一变量。OECD对14国地区迁移的研究发现，更好的收入与就业会吸引人口，而高房价会显著阻碍流动，且不同国家的反应差异很大。对中国流动人口的研究也显示，住房成本、医疗、教育、交通、社会包容度与家庭负担都会影响定居意愿，而不是由名义工资单独决定。PLOS ONE研究

家庭网络同样是一种经济基础设施。亲属可以提供照护、信息、临时住房与情感支持，这些没有完全进入GDP，却会显著改变一个城市的实际生活成本；既有移民网络也能帮助新来者寻找工作、交通和本地资源。澳大利亚迁移研究同时提醒我们，行政安排可以影响最初去向，却未必保证长期留存。

因此，当传统中心的收入和职业溢价下降，而住房、税收、竞争、身份政策与情感成本上升时，人们的选择可能改变。改变不必表现为“回家”：它可能是从超级城市转向区域中心，从永久定居转向阶段性居住，从单一国别下注转向多节点职业路径，也可能是远程工作与频繁往返。流动没有消失，只是从单向攀登变成了组合配置。

二、全球化没有终结，而是在分层重组

“去全球化”常被当作一个单一趋势，但贸易、资本、供应链和人才流动并不共用一只时钟。全球货物贸易正在沿地缘政治边界重新布线，资本更重视合规和韧性，人才政策则在短缺技能、家庭团聚、留学与人道主义通道之间出现截然不同的方向。把这些变化压缩成“全球化终结”，会误导企业的资源配置，也会让政策滑向过度反应。

流量数据则说明“分化”比“下降”更准确。OECD永久型移民从2023年的650万降至2024年的620万，同比减少4%，却仍比2019年高15%；同期约230万份新的临时外籍劳工许可与2023年大体持平，比2019年高26%，而新国际高等教育学生流入下降13%至180多万，人道主义移民和庇护申请反而上升。OECD 2025年《国际移民展望》表明，劳动、家庭、学生和人道主义迁移不能被画在同一条趋势线上。

政策也不是简单的“全部关门”。加拿大转向数量管理，英国显著提高技术工作签证门槛，美国增加审查与项目完整性要求；与此同时，欧盟仍在建设面向短缺职业的EU Talent Pool，澳大利亚则试图把技术移民路径与生产率和劳动力需求更紧密地连接。欧盟委员会与澳大利亚内政部《Migration Strategy》所反映的，不是开放与封闭的二元选择，而是各国对规模、技能和身份路径的重新排序。

美国H-1B数据尤其说明口径的重要。FY2026合格电子注册数为343,981份，较FY2025的470,342份下降26.9%，但这不是正式申请量下降，也不能单独证明高技能需求同比萎缩；受益人中心抽签与反重复注册措施改变了注册行为。美国公民及移民服务局（USCIS）的数据需要被准确称为“合格电子注册”，而不是渲染成移民体系已经关闭。

同样需要澄清的是，美国2025年公告曾对部分新H-1B申请设置10万美元附加费，但截至2026年9月9日，相关实施指引已被联邦地区法院撤销，上诉法院也拒绝暂缓该裁定，DHS表示遵守法院命令。白宫公告记录了政策起点，USCIS H-1B页面则记录了诉讼后的阻断状态。EB-1A与NIW的46.6%和64.3%也只是NFAP对USCIS数据所作的FY2025第四季度拒签率分析，不是FY2025全年拒签率。Forbes对该季度分析的报道

贸易全球化同样是重组，而非消失。WTO指出，美中双边贸易自2018年以来相对各自与世界其他地区的贸易慢约30%，2022年后假设地缘集团之间的货物贸易也比集团内部慢4%—6%；但2023年实际货物贸易仍比疫情前峰值高6%、比2015年高19%。WTO关于“再全球化”的分析更接近“网络重新布线”的判断。

安全与韧性问题不能被轻率否定。关键药品、半导体、能源与基础设施确有集中度风险，国家有责任评估极端情景；问题在于，韧性不等于不计成本的自给自足，安全审查也不应无限扩张为所有领域的常态壁垒。IMF的情景模拟显示，地缘经济碎片化的长期损失取决于分裂程度与调整成本，可能从有限脱钩下约0.3%的全球GDP，到严重集团化和高调整成本下的更大损失；这些是风险情景，不是既成事实。IMF《The Costs of Geoeconomic Fragmentation》

三、中国城市地图的分化：一线放缓，不等于集体撤离；强二线承接，不等于普遍繁荣

中国城市人口变化最容易被两个叙事误读：一是“一线城市大逃离”，二是“强二线全面胜出”。2020年至2025年，北京常住人口约减少9.31万，上海约减少1.68万，确实表现为小幅下降或基本横盘；但广州约增加42.44万，深圳约增加68.84万，仍保持增长。北京的2020年七普与2025年公报、上海的2020年七普与2025年公报、广州的2020年七普与2025年公报，以及深圳的2020年七普与2025年公报说明，一线城市内部首先发生的是分化，而不是同步撤退。

强二线和新一线城市总体上承接了更多人口，但端点增量不能自动等同于净迁入，更不能直接证明新增人口都是高技能人才。杭州2020年至2025年常住人口约增加76.4万；合肥约增加63.5万，但合肥的数字是常住人口总量变化，不是五年净流入，且其2025年人口仅比上一年增加0.3万。杭州人口与经济信息和合肥人口与产业报道共同提示我们：承接是真实的，速度却会显著变化。

更重要的是，这些城市不是同一种产品。杭州的吸引力来自数字经济、平台生态、软件与信息服务及创新网络；合肥更依赖新能源汽车、集成电路、新型显示和先进制造形成的硬科技岗位。杭州2025年信息显示其营利性服务业与研发投入保持较强基础，合肥相关数据则显示电子信息制造与新能源汽车产量的产业支撑。

成都的组合是消费市场、科技服务、电子信息、装备制造与宜居性的共同作用；武汉依靠高校科研、光电子、汽车和高技术制造形成知识与产业的接口。成都统计年鉴2025记录了其研发投入与产业结构，武汉2025年统计公报则显示高技术制造业占规上工业增加值26.2%，并拥有41个全国重点实验室。

西安的比较优势更多来自航空航天、科研机构、高技术制造与工程人才；苏州则凭借先进制造、外资研发网络和长三角协同，把上海的市场、资本与国际连接同本地产业链相结合。西安2025年数据与苏州2025年统计公报显示，两座城市的承接逻辑都建立在具体产业平台上，而不是抽象的“低成本”。

人口增长也不等于普遍繁荣。一个城市可能吸引大量毕业生，却无法提供足够的中高阶岗位；也可能拥有亮眼产业产值，却因住房、教育、医疗或职业天花板而留不住家庭。投资者真正需要跟踪的不是一个总人口数字，而是25—44岁人口、毕业生留存、岗位薪酬、研发人员、企业招聘、住房负担、通勤效率、风险资本和产业订单的组合。

四、人才回流的双刃剑：从人口数字到生产率，需要一座“制度桥梁”

人才流动能够提高创新能力，但人才只是必要投入，不是自动兑现的红利。美国研究显示，移民发明人约贡献研究期内23%的专利，计入与本土发明人的合作后，移民直接或间接关联36%的美国专利产出；这个结果说明跨境人才扩大知识组合与协作网络，却不意味着每位移民都会产生同等创新。NBER研究摘要

对中国上市公司的研究提供了类似的条件性结论。对2,499家上市公司、2002—2016年数据的分析发现，归国员工和技术人员与专利增长正相关，海外工作经历的作用强于单纯留学经历，但效果因所有制、区位与企业规模而异。Science and Public Policy研究摘要因此，“回来了多少人”只能是起点，不能代替“形成了多少可扩散的知识与生产率”。

这座制度桥梁至少有七个桥墩：真实且持续的岗位需求，能够容纳试错的科研与工程平台，具备吸收能力的企业，连接实验室与市场的融资渠道，可信的知识产权与商业规则，可负担的住房和教育医疗，以及对不同背景人才开放的组织文化。任何一个关键桥墩缺失，都可能让人才困在低匹配岗位中。

短期内，人才下沉甚至可能加剧高学历就业拥挤。地方岗位数量、层级与专业结构如果跟不上，学历会成为筛选工具而不是生产力工具，工资也可能被供给增加压低；补贴到期后，人才还可能再次离开。城市需要警惕把“招来”误当成“用好”，企业也需要避免把国际履历当作装饰性标签。

这也解释了为什么人才政策必须从个人扩展到家庭。住房券可以降低进入成本，却不能替代伴侣职业、子女入学、医疗可及与长期预期；一次性创业补贴可以启动项目，却不能替代稳定融资、采购机会和公平退出机制。真正有效的政策，不是把人才“安排”在某个名册里，而是允许其在透明规则下创造、流动和失败后再来。

五、国家竞争的新前线：以内部竞争力实施“进攻性防御”

在地缘竞争上，防御当然有其位置。国家必须保护关键基础设施、限制明确的安全漏洞，并为极端冲击准备冗余；但如果防御主要表现为不断扩大禁令、抬高普遍性壁垒和减少交流，它最终可能同时削弱对手与自己。真正可持续的“进攻性防御”，是让本国成为科研、创业、制造和生活更有吸引力的地方。

这意味着把竞争焦点从“阻止别人获得什么”转向“自己能创造什么”。基础研究、大学、职业教育、数字和交通基础设施、稳定能源、资本市场、营商环境与法治，都是国家安全的生产性底座；吸引全球人才并让其与本地人才协同，则是把这些投入连接起来的机制。边境政策可以筛选风险，却无法替代实验室、学校、企业和城市治理。

开放并不要求天真。对敏感技术可以实行窄而清晰、可复核的限制，对供应链可以建立多来源与库存，对外资可以设置透明审查；但政策边界越模糊，企业越会推迟投资，人才越会为不确定性索取风险溢价。安全政策的质量，应以它在降低具体风险的同时，是否尽量保留创新、贸易和人才连接来衡量。

高技能人才对发达经济体仍至关重要。2021年，外国出生者约占美国STEM劳动力的19%；在科学与工程职业中，博士持有者的外国出生比例达到43%。美国国家科学基金会/NCSES说明，政策摩擦可能侵蚀吸引力，但不能据此断言美国或其他传统创新中心的高科技能力已经不可逆转地衰败。

六、战后秩序的历史镜鉴：开放、合作与制度化共同利益

二战后的制度建设值得回看，不是因为那是一个没有冲突的黄金时代，而是因为当时的决策者尝试把共同利益写进可持续的安排。1947年提出、1948年立法实施的马歇尔计划向西欧提供超过120亿美元援助，帮助重建并扩大市场，同时也服务于美国的地缘战略；其精确增长贡献至今仍有争论。美国国务院历史办公室

1944年布雷顿森林会议由44国代表设计战后货币与重建合作框架，并产生IMF与国际复兴开发银行章程；它的核心是金融稳定与重建，而不是一般意义上的自由贸易协议。世界银行档案1947年由23个创始缔约方签署的GATT，则以45,000项关税减让覆盖约当时全球五分之一贸易，并在后来多轮谈判中推动降低壁垒，但它在农业、纺织、服务、投资与争端解决方面都有明显缺口。WTO历史

1950年的《舒曼宣言》进一步提供了制度设计的启示：法国、西德、意大利、荷兰、比利时和卢森堡把煤钢生产置于共同高级机构之下，让关键军工投入受到共同约束。欧盟《舒曼宣言》历史资料它的价值不在于证明贸易自动带来和平，而在于说明相互依赖只有被规则、机构与政治承诺管理，才能降低零和竞争的空间。

历史同样警告我们不要使用单因果叙事。1930年《斯穆特—霍利关税法》恶化了贸易关系并引发报复，但大萧条在法案生效前已经开始，货币、银行、金本位和债务通缩都是更广泛的解释；关税不能被写成大萧条或二战的唯一原因。NBER历史研究同理，今天的地缘冲突也不能被简化为贸易壁垒的机械结果。

七、企业、城市与个人的战略含义

对企业：从“全球布局”转向“多节点人才组合”

企业不应以“全球化结束”为基准情景，也不应假设旧有中心会永远占优。研发、生产、销售、数据与合规岗位的可迁移性不同，最合理的组织可能不是把所有职能搬到一个低成本城市，而是在多个节点之间配置团队，并为签证、贸易和供应链变化保留转换能力。

选址时，工资只是最显眼的成本，不一定是最重要的成本。企业应同时评估人才密度与流失率、供应商和客户距离、大学与实验室、住房和通勤、家属安置、数据规则、融资渠道与政策可预期性；只有当这些因素共同成立，名义成本优势才会变成生产率优势。

对城市：从“抢人”转向“用人、留人、成就人”

城市竞争的下一个阶段，不是继续加码一次性补贴，而是建设完整的职业阶梯和生活周期。引进青年毕业生之后，要有中层岗位；引进科学家之后，要有稳定团队与设备；引进创业者之后，要有首批客户、长期资本和可预期的退出规则。

城市还需要公开衡量人才政策的真实结果。除新增常住人口外，应跟踪三年和五年留存率、专业与岗位匹配度、家庭定居、研发产出、企业成长以及公共服务承载，并允许结果反向纠正政策。透明评估本身就是一种吸引力，因为人才看重的不只是优惠，也看重规则是否可信。

对个人与家庭：不要追随口号，要管理机会净值与可选性

个人结论不是“只能回国”或“应该回乡”。更稳健的做法，是比较税后收入、职业学习曲线、住房和教育医疗、身份路径、伴侣机会、家庭照护与社会归属，并为这些变量设置自己的权重；一个对单身创业者极具吸引力的城市，可能并不适合需要照顾两代人的家庭。

归属不应被浪漫化，也不应被低估。家庭网络可以降低成本，熟悉文化可以提升生活质量，但如果家乡缺少匹配岗位、规则和专业网络，回去也可能损害长期发展；反之，只要海外或一线中心仍能提供足够的机会净值，留下仍然是理性选择。迁徙是一项具体计算，而不是身份忠诚测试。

结论：迁徙不会消失，目的地的等级秩序正在改变

当传统中心的机会溢价下降，而生活与制度成本上升，人们并不会停止寻找更好的生活。他们会改换方向、缩短周期、增加节点、重新依靠家庭网络，或者在几个制度与产业空间之间循环。全球迁移的下一阶段，可能更少由单一“梦想目的地”定义，更多由多中心的机会组合定义。

对国家而言，竞争的焦点将从阻止人才离开或阻止对手获得人才，转向创造值得人才留下的制度环境。对企业而言，优势来自跨节点组织能力；对城市而言，优势来自人才转化能力；对个人而言，优势来自看清机会净值并保留选择权。

迁徙并未走到终点。正在松动的，是少数传统中心对机会、身份与归属的垄断；正在形成的，是一张更分散、更区域化、更循环，也更需要精细判断的新地图。

参考资料

以下仅列正文实际引用的来源，链接均为完整URL；访问与事实核验截止于2026年9月9日。

When opportunity premiums fall and the costs of place rise, mobility does not end—it changes direction

Executive Summary

Migration is often narrated as a journey from the periphery to the center: from a smaller city to a metropolis, or from an emerging economy to a rich one. That framing is increasingly incomplete. For individuals and families, mobility is a long-duration allocation decision involving not only wages, but also career progression, housing, taxation, education, healthcare, legal status, family care, professional networks, and the possibility of belonging.

That bundle is now being repriced. When the income, career, and status premium offered by an established center no longer compensates for its financial, institutional, and emotional costs, people do not necessarily stop moving. They redirect—to regional hubs, third countries, hometowns, lower-cost innovation centers, or a portfolio of locations connected through circular and temporary mobility.

The data do not support the claim that the era of global migration is ending. The world hosted an estimated 304 million international migrants in 2024, or 3.7% of the global population, although growth in the stock slowed to less than 1% annually between 2020 and 2024.UN DESA, International Migrant Stock 2024 The more defensible conclusion is that migration remains historically high but is becoming slower, more differentiated by category, more regional, and more circular.

The strategic question is therefore not simply who can attract talent, but who can use it. Jobs, industrial clusters, research platforms, finance, affordable housing, schools, healthcare, predictable rules, and an open culture form the institutional bridge between a larger talent pool and higher productivity. Without that bridge, return migration or domestic decentralization may produce credential congestion rather than innovation.

Introduction: Not a Great Homecoming, but a Recalculation of Opportunity

Some return to their place of origin. Others choose a third country, a regional hub, a temporary assignment, or a life divided between multiple places.

A more useful question is whether the expected net value of a location still exceeds the total price a migrant must pay for it. For decades, a small group of global cities and advanced economies offered a powerful center premium: higher salaries, denser professional networks, stronger universities, deeper capital markets, and more credible pathways to status.

China is therefore best understood as an instructive mirror rather than a global template. Its urban redistribution helps reveal a wider mechanism: when the top destination no longer offers a sufficiently large premium, mobility becomes multi-directional. It does not prove that international migration will follow the same routes or that all migrants will return to their countries of birth.

I. The Mobility Equation: Opportunity Premium Minus the Full Cost of Place

The decision to move can be expressed as a practical, if necessarily imperfect, balance sheet. Expected net value equals wages and career upside, learning and innovation networks, public services, family benefits, legal security, and quality of life—minus housing, commuting, relocation, information, adaptation, family separation, and policy risk. The weights vary sharply by life stage and occupation.

Income and employment remain central. Yet an OECD gravity model covering regions in 14 countries found that people are attracted by better income and labor-market prospects while high housing prices impede movement, with substantial differences across countries. A large study of China’s floating population similarly linked settlement intentions to housing costs, healthcare, education, transportation, social openness, and family circumstances—not wages alone.

Family and community networks are part of the economic infrastructure of mobility. Relatives can provide childcare, eldercare, information, temporary housing, and emotional support, all of which lower the real cost of living even when they do not appear in conventional income statistics. Research on migrants in Australia also shows that established networks can facilitate transport, accommodation, employment, and access to local resources, while visa rules may shape an initial destination without guaranteeing long-term retention.Journal of Population Research

This framework explains why mobility can reverse direction without becoming irrational. A move from a global city to a regional center may reduce nominal salary while raising disposable income, family support, time, and the probability of home ownership. Conversely, remaining in an expensive center can still be rational when the learning curve, professional network, capital access, or status pathway remains unusually valuable.

II. Globalization Is Not Ending; Its Layers Are Being Rewired at Different Speeds

The stock of international migrants demonstrates the persistence of cross-border life. It reached approximately 304 million in 2024, almost double the 153.9 million recorded in 1990; the global population share rose more modestly, from 2.9% to 3.7%. Growth slowed to less than 1% a year from 2020 to 2024, making “high-level deceleration” a more accurate description than “termination.”UN DESA

Annual flows reveal divergence within that total. Permanent-type migration to OECD countries fell from a record 6.5 million in 2023 to 6.2 million in 2024, a 4% decline, but remained 15% above 2019. Roughly 2.3 million new temporary foreign-worker permits or authorizations were issued in 2024 excluding Poland, broadly unchanged from 2023 and 26% above 2019; new international tertiary-student inflows fell 13% to more than 1.8 million, while humanitarian migration rose 23% and asylum applications increased 13% to 3.1 million.OECD, International Migration Outlook 2025

The composition of permanent migration further complicates any simple labor-market story. Family-related migration accounted for 44% of OECD permanent inflows in 2024, compared with 17% for labor, 13% for humanitarian channels, 19% for free movement, and 7% for other categories.OECD Family networks and humanitarian obligations do not respond to business cycles in the same way as employer-sponsored recruitment.

Nor are advanced economies uniformly closing their doors. Canada has shifted from rapid expansion toward tighter quantity management; the United Kingdom has raised salary and skill thresholds; the United States has increased screening and changed program-integrity rules. At the same time, the European Union is developing an EU Talent Pool for shortage occupations, and Australia’s migration strategy seeks clearer links between skilled pathways, labor demand, and productivity.European Commission Australian Department of Home Affairs

The United States illustrates why precise definitions matter. Eligible H-1B electronic registrations fell 26.9%, from 470,342 for FY2025 to 343,981 for FY2026, but that is not a 26.9% decline in filed petitions or proven employer demand. The beneficiary-centric selection process and anti-duplication measures materially changed registration behavior.USCIS H-1B Electronic Registration Process

Other prominent figures require equally careful qualification. A September 2025 presidential proclamation imposed a $100,000 supplemental payment on certain new H-1B petitions, but by September 9, 2026, implementation guidance had been vacated by a federal district court, the First Circuit had denied a stay on July 24, and DHS stated that it was complying with the court order.White House proclamation USCIS H-1B page The widely cited 46.6% EB-1A and 64.3% NIW denial figures refer to the fourth quarter of FY2025 in an NFAP analysis of USCIS data—not full-year rates.Forbes

Trade tells a parallel story of rewiring. The WTO reported that US–China bilateral trade had grown 30% more slowly since 2018 than each country’s trade with the rest of the world, and that merchandise trade between hypothetical geopolitical blocs had grown 4%–6% more slowly than trade within blocs since 2022. Yet real merchandise trade in 2023 remained 6% above its pre-pandemic peak and 19% above 2015.WTO analysis on re-globalization

Supply chains are also shifting without cleanly separating. World Bank research on 2017–2023 trade patterns found that US and European sourcing moved toward economies including Vietnam, India, and Mexico, but value chains remained intertwined; infrastructure, logistics, skills, finance, ICT, and research capacity shaped who could capture relocation.World Bank research “Friend-shoring” may change the map, but it does not eliminate interdependence.

National-security and resilience concerns should not be dismissed. Concentrated dependence in medicines, semiconductors, energy, or critical infrastructure can create real vulnerabilities. The policy challenge is to distinguish targeted resilience from open-ended insulation: diversification and redundancy can reduce risk, while indiscriminate barriers can raise costs, slow diffusion, and erode the innovation base they are meant to protect.

III. China’s Differentiated Urban Map: First-Tier Moderation Is Not a General Exodus

China’s urban debate is often organized around two sweeping claims: that residents are fleeing the first tier, and that strong second-tier cities are uniformly taking over. Neither survives a close look at the population data. Between the 2020 census benchmark and year-end 2025, Beijing’s permanent population declined by roughly 93,100 and Shanghai’s by about 16,800, but Guangzhou added approximately 424,400 residents and Shenzhen about 688,400.Beijing 2020 census Beijing 2025 communiqué Shanghai 2020 census Shanghai 2025 communiqué Guangzhou 2020 census Guangzhou 2025 communiqué Shenzhen 2020 census Shenzhen 2025 communiqué

Strong regional centers generally gained population, but their trajectories also differ. Hangzhou’s permanent population increased by about 764,000 between 2020 and 2025. Hefei’s rose by roughly 635,000 over the same endpoints, but that is a total population increase—not synonymous with net migration—and its year-on-year increase slowed to just 3,000 in 2025.Hangzhou 2025 population and economic information Hefei population and industry report

These cities are not interchangeable cost havens. Hangzhou combines platform businesses, software and information services, digital commerce, and a dense innovation ecosystem. Its 2025 data point to continued strength in profit-oriented services and an estimated R&D intensity of about 4%, offering a concrete base for digital and entrepreneurial talent rather than merely cheaper housing.Hangzhou city information

Hefei’s proposition is different. Its attraction rests on advanced manufacturing and hard technology, including new-energy vehicles, semiconductors, and next-generation displays; in 2025, value added in electronic-information manufacturing rose 60.6%, automobile manufacturing grew 11.1%, and new-energy vehicle output reached 1.371 million units.Hefei population and industry report The sharp slowdown in population growth that same year is a useful warning: strong industrial output does not guarantee continuously accelerating demographic gains.

Chengdu offers a mixed economy of consumption, technology services, electronic information, equipment, and healthcare, reinforced by a large regional market and substantial R&D spending. Its 2024 R&D expenditure reached RMB 92.089 billion, equal to 3.92% of GDP, while electronic information, equipment, and health industries provide platforms for skilled employment.Chengdu Statistical Yearbook 2025

Wuhan’s comparative advantage lies in the interface between education, research, optoelectronics, automobiles, and high-technology manufacturing. In 2025, high-tech manufacturing value added grew 16.6% and accounted for 26.2% of above-scale industrial value added; the city reported 41 national key laboratories and 92 members of the Chinese academies.Wuhan 2025 statistical communiqué Those institutions create absorptive capacity, although Wuhan’s population growth cannot be assumed to consist entirely of high-skilled inflows.

Xi’an is anchored by aerospace, defense-related research, engineering universities, and high-tech manufacturing. Its 2025 data show 8.4% growth in aerospace and other transport-equipment manufacturing and 15.4% growth in high-tech manufacturing investment, alongside a deep base of research institutions and technical enterprises.Xi’an 2025 statistical communiqué coverage The fit is strongest for specialized engineering and research-intensive activity, not for every mobile professional.

Suzhou’s advantage is its advanced-manufacturing depth and integration with the Yangtze River Delta. High-tech manufacturing output reached RMB 1.85888 trillion in 2025 and contributed 53.3% of the increase in above-scale industrial output; the city also reported a talent pool of 4.25 million and 283,000 university graduates attracted or retained during the year.Suzhou 2025 statistical communiqué Its proximity to Shanghai matters because regional complementarity—not isolation from the first tier—is part of the model.

Population growth is therefore an outcome indicator, not proof of broad prosperity or talent quality. A city can attract graduates without creating enough mid-career positions, or expand industrial output while remaining difficult for families to settle in. Executives and investors should supplement headline population totals with age structure, graduate retention, job vacancies, R&D employment, housing burden, commuting time, venture funding, and industrial orders.

IV. The Double Edge of Talent Return: Building the Institutional Bridge to Productivity

Talent can increase innovation, but it does not operate in a vacuum. Research on the United States estimates that immigrant inventors accounted for about 23% of patents during the study period and were directly or indirectly connected to 36% of US patent output when collaboration with native-born inventors is included.NBER research summary The finding supports the value of diverse knowledge and networks; it does not imply that every migrant produces innovation or that headcount alone predicts output.

Evidence from China reaches a similarly conditional conclusion. A study of 2,499 listed companies from 2002 to 2016 linked returnee employees and technical personnel to patent growth, with overseas work experience showing a stronger effect than study alone; the results varied by ownership, location, and company size.Science and Public Policy research summary Return is thus an input into a chain of conversion, not a synonym for industrial upgrading.

That conversion requires an institutional bridge. The first pillar is sustained demand: firms must have products, orders, and technical problems worthy of the available skills. The second is an operational platform—laboratories, engineering teams, data, equipment, and management capable of integrating specialized knowledge.

The third pillar is finance that can carry a project from experimentation through commercialization, rather than rewarding only recruitment or initial formation. The fourth is a credible framework for intellectual property, contracts, procurement, and market access. The fifth is the household environment: housing, schooling, healthcare, spousal employment, and a stable expectation of belonging.

When the bridge is incomplete, talent inflows can deepen credential congestion. A city may recruit large numbers of degree-holders into a labor market with too few high-productivity roles, turning education into a screening device rather than a productive asset. Wage compression, underemployment, exam competition, and subsequent departure can follow.

This is why talent policy must graduate from “recruitment” to “retention and deployment.” A housing grant can reduce entry costs but cannot substitute for a spouse’s career or a child’s school. A startup subsidy can launch a company but cannot replace patient capital, a first customer, transparent procurement, or a predictable exit.

V. The New Front Line of National Competition: Offensive Defense Through Domestic Capability

A more durable strategy is “offensive defense” through internal competitiveness. Its purpose is not military aggression or economic coercion, but the creation of such strong research institutions, schools, infrastructure, companies, markets, and living conditions that talent and investment choose to participate. The most effective protection is a system capable of learning, adapting, and producing at the frontier.

Openness should not mean strategic innocence. Controls on genuinely sensitive technologies can be narrow, explicit, reviewable, and coordinated with investment in alternatives. Supply-chain resilience can emphasize multiple sources, interoperability, inventories, and trusted standards rather than universal domestic substitution.

The continuing importance of international talent to advanced economies is measurable. In 2021, approximately 7.02 million US STEM workers were foreign-born, representing 19% of the STEM workforce; among people in science and engineering occupations, the foreign-born share reached 43% for doctorate holders, 37% for master’s holders, and 19% for bachelor’s holders.NSF/NCSES Policy friction can erode this advantage, but it does not justify a claim that US high technology is irreversibly finished.

Talent competition is therefore a dynamic contest in institutional quality. Walls can alter short-term flows. The long-term map will be determined by where people can do consequential work, build durable lives, and connect their knowledge to organizations capable of using it.

VI. The Postwar Lesson: Institutionalizing Shared Interests Without Mythologizing Openness

The postwar order is relevant not because it eliminated rivalry, but because its architects attempted to place shared gains inside institutions. The Marshall Plan, proposed in 1947 and authorized in 1948, provided more than $12 billion to support Western European reconstruction and expanded markets for US goods, while also serving a strategic objective of containing communism; historians still debate the program’s precise contribution to growth.US Department of State, Office of the Historian

At Bretton Woods in July 1944, delegates from 44 countries designed a framework for postwar monetary and reconstruction cooperation that produced the charters of the IMF and the International Bank for Reconstruction and Development. Its primary objectives were stability, reconstruction, and long-term development—not a general free-trade agreement.World Bank archives

The General Agreement on Tariffs and Trade was signed in 1947 by 23 founding parties. Its 45,000 tariff concessions covered roughly one-fifth of world trade at the time and subsequent rounds reduced barriers, but the system remained incomplete in agriculture, textiles, services, investment, and dispute settlement before the WTO replaced it.WTO history

The 1950 Schuman Declaration supplied another institutional insight. France, West Germany, Italy, the Netherlands, Belgium, and Luxembourg placed coal and steel production under a common High Authority, jointly constraining inputs central to war-making.European Union history of the Schuman Declaration Its lesson is not that commerce automatically creates peace, but that interdependence becomes stabilizing when rules, institutions, and political commitments manage it.

History also cautions against monocausal analogies. The 1930 Smoot–Hawley tariff worsened trade relations and prompted retaliation, but the downturn had begun before the law took effect, and monetary collapse, banking crises, the gold standard, and debt deflation figure centrally in explanations of the Great Depression.NBER historical research Neither the Depression nor the Second World War can be attributed to one tariff statute.

The usable lesson is more modest and more demanding. Open institutions can reduce friction, increase predictability, enlarge shared interests, and raise the cost of conflict. Peace also depends on security arrangements, domestic distribution, macroeconomic stability, political restraint, and credible enforcement.

VII. Strategic Implications for Companies, Cities, and Individuals

For Companies: Build a Multi-Node Talent Portfolio

Companies should not plan around the assumption that globalization has ended, nor around the opposite assumption that the old centers will remain unchallenged. Research, production, sales, regulatory, and data functions have different mobility constraints. A resilient operating model may distribute them across several nodes rather than move an entire organization to the location with the lowest headline cost.

Location analysis must look beyond wages. Talent depth and turnover, universities and laboratories, suppliers and customers, housing and commuting, spousal employment, schools, data regulation, finance, and policy predictability all influence effective productivity. A cheaper payroll can become an expensive mistake if the firm cannot retain experienced people or connect them to a viable ecosystem.

China’s regional centers illustrate the case for functional matching. Hangzhou suits platform and digital-business capabilities; Hefei offers hard-tech manufacturing demand; Chengdu combines technology services and a major consumer market; Wuhan connects research to optoelectronics and industrial activity; Xi’an specializes in aerospace and engineering; Suzhou links advanced manufacturing to the wider Yangtze River Delta. The objective is not to choose the next fashionable city, but to align each activity with the location that can compound it.

For Cities: Move From Recruiting People to Enabling Careers and Lives

The next phase of urban talent competition will not be won by the largest one-time subsidy. Cities need a complete career ladder and household life cycle: entry-level opportunity, mid-career advancement, research teams, equipment, customers, capital, quality schools, healthcare, affordable housing, and credible commercial rules.

Recruiting a scientist without a laboratory, an entrepreneur without a first customer, or a graduate without a path to advancement produces visible intake but weak retention. Policy should be evaluated through three- and five-year retention, job matching, family settlement, firm formation, R&D outcomes, wage progression, and spillovers to local workers—not only through the number of people registered in a program.

The winning city is not necessarily the cheapest one. It is the place where the net value proposition is legible: people understand why they should come, how they can progress, what happens if a venture fails, and whether their families can remain. Predictability turns a temporary incentive into durable trust.

For Individuals and Families: Optimize Net Opportunity and Preserve Optionality

The personal conclusion is not “go home” or “leave the first tier.” Individuals should compare after-tax income, the slope of career learning, housing, education and healthcare, legal-status pathways, a partner’s prospects, family care, and social belonging. The correct weights are personal, and they change over time.

Mobility also need not be an irreversible bet. Short assignments, trial periods, two-city arrangements, regional moves, remote collaboration, and staged education can produce information before a permanent commitment. Visas, savings, portable skills, language ability, and cross-regional networks should be treated as option assets.

Home networks can lower costs and strengthen well-being, but home is not automatically the best professional platform. A return to a location without suitable jobs, transparent rules, or peer networks may reduce long-term opportunity. Conversely, remaining abroad or in a first-tier city remains rational when the opportunity premium still compensates for the full cost.

Conclusion: Mobility Will Persist, but the Hierarchy of Destinations Is Changing

When the opportunity premium of a traditional center falls and the financial, institutional, or emotional cost of place rises, people do not stop seeking better lives. They change destinations, shorten commitments, rely more heavily on family networks, add nodes, or circulate among several systems. The next era of migration may be defined less by a handful of dream destinations than by portfolios of opportunity.

For nations, the contest is shifting from preventing people from leaving—or preventing rivals from gaining them—to building institutions worth joining. For companies, advantage lies in organizing talent across nodes. For cities, it lies in converting human capital into productive capability. For individuals, it lies in understanding net opportunity and preserving the capacity to choose.

Migration is not approaching an endpoint. What is weakening is the monopoly that a small set of traditional centers once held over opportunity, status, and belonging. In its place is emerging a more distributed, regional, circular, and demanding map—one that rewards institutions capable not merely of attracting talent, but of enabling it to contribute and remain.

Sources

以下仅列正文实际引用的来源，链接均为完整URL；访问与事实核验截止于2026年9月9日。