移民政策能否像操作系统一样运行？新加坡 COMPASS 与持续校准的 85% 艺术

Can Immigration Policy Run Like an Operating System? Singapore's COMPASS and the 85% Art of Continuous Calibration

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

核心问题 · Core Problem: 多数国家把移民参数一次性设定后便在政治僵局中冻结，任由工资、住房与劳动力市场脱节数十年。其结果正是澳大利亚（有筛选无闸门）与德国（有设计无执行）展示过的失败模式：参数与现实持续背离，直至体系失去合法性。 Most countries set immigration parameters once and then freeze them under political deadlock, letting wages, housing, and labor markets drift out of alignment for decades. The result is the failure mode seen in Australia (selection without a gate) and Germany (design without execution): the parameter and the reality diverge until the system loses legitimacy.
理论解法 · Theoretical Solution: 新加坡把移民当作一套持续校准的操作系统来运行，分三层：就业准证薪资闸门按行业与年龄锚定本地 PMET 工资前三分之一（2025 年 5,600–6,200 新元，2027 年升至 6,000–6,600 新元）；COMPASS 计分框架要求在薪资、学历、多元化、本地就业四个维度拿满 40 分，把企业层面的外部性重新定价回雇主；在工作准证层，外劳税（每月 650 新元）与比例配额（服务业 10%）把承载力约束变成硬工程参数。 Singapore runs immigration as a continuously calibrated operating system in three layers: an Employment Pass salary gate benchmarked to the top one-third of local PMET wages by sector and age (S$5,600–S$6,200 in 2025 rising to S$6,000–S$6,600 by 2027); the COMPASS points framework requiring 40 points across salary, qualifications, diversity, and local-employment criteria, which prices firm-level externalities back onto employers; and a work-permit layer where levies (S$650 per month) and dependency ratio ceilings (10% in services) make the carrying-capacity constraint a hard engineering parameter.
实证数据 · Empirical Data Metric: 关键校准参数：就业准证合格薪资自 2025 年起为一般行业 5,600 新元、金融业 6,200 新元，2027 年升至 6,000 / 6,600 新元，并随年龄递增，45 岁以上申请人约达 10,700 新元；COMPASS 及格线 40 分，薪资项按同行业同年龄段 PMET 工资第 65–90 分位计分；S 准证薪资底线 3,300 新元（2027 年升至 3,600 新元），外劳税统一为每月 650 新元；比例配额为服务业 10%、建筑/制造/海事/加工业 15%；非居民约占总人口 30%。 Key calibration parameters: EP qualifying salary S$5,600 (general) / S$6,200 (financial services) from 2025, rising to S$6,000 / S$6,600 by 2027 and scaling with age to roughly S$10,700 for candidates aged 45+; COMPASS pass mark of 40 points with salary scored against the 65th–90th percentile of sector-age PMET wages; S Pass salary floor S$3,300 rising to S$3,600 by 2027 with the levy unified at S$650 per month; dependency ratio ceilings of 10% (services) and 15% (construction, manufacturing, marine, process); non-residents constitute roughly 30% of the total population.
核心观点 · Key Takeaway: 澳大利亚造出了精密筛选机器却没有总量闸门，德国设计了 A− 的机器却只有 C+ 的执行。新加坡做了一件性质完全不同的事：把移民政策变成一套持续校准的操作系统——与本地工资挂钩的薪资闸门、把企业外部性定价的 40 分 COMPASS 计分制、以及用外劳税加比例配额把承载力变成工程参数而非政治口号。本篇为《精细化移民治理》系列第 3 篇，考察锚文「85% 安全系数」在现实中的最近似实现，并追问大国能复制什么、不能复制什么。 Australia built a superb selection machine but no intake gate. Germany designed an A-minus machine executed at C-plus. Singapore did something categorically different: it turned immigration into a continuously calibrated operating system — wage-indexed salary gates, a 40-point COMPASS score that prices firm-level externalities, and levies plus ratio caps that make carrying capacity an engineering parameter rather than a political slogan. Part 3 of the Precision Immigration Governance series examines the closest real-world implementation of the anchor framework's 85% safety coefficient — and asks what larger states can and cannot copy.
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

作者：殷彤博士 · 美国洞见桥全球公司创始人 · 2026 年 8 月 · 《精细化移民治理》系列第 3 篇，前接锚文、第 1 篇：澳大利亚与第 2 篇：德国。

本系列第 1 篇指出，澳大利亚造出了英语世界最精密的筛选机器，却在二十年里几乎从未校准过总量闸门；第 2 篇指出，德国在半个世纪的否认之后设计出了一台 A− 的机器，却被自己的行政体系执行成 C+。两种失败看似不同，根子是同一个：参数被一次性设定后便任其漂移，而经济早已走远。第 3 篇转向唯一一个从一开始就拒绝这种失败模式的国家。据我所知，新加坡是全球唯一像央行管理货币政策一样管理移民政策的政府——把它当作一套持续校准的操作系统，而非反复爆发的政治危机。它也是锚文核心处方在现实中的最近似实现：让体系贴着可测量的承载力运行，绝不越线，并在裂缝变成怨气之前完成再校准。

为什么只有新加坡把移民当作操作系统来运营？

锚文提出，最优移民政策应当是规则化的、逆周期的、高频再校准的——更接近通胀目标制而非立法。新加坡是唯一真正把它造出来的主要经济体。其结构性条件部分独一无二：约六百万人口的城市国家，没有内陆地区，没有联邦式否决者，住房供给由国家主导（组屋覆盖约八成居民），人力部拥有真正的技术官僚授权。非居民约占总人口三成——这个比例在几乎任何其他国家都会引发政治断裂，而在新加坡它被当作一个运营变量来管理。但更深层的原因是理念层面的。自 1990 年代起，人力部就把外籍劳动力视为一个带可测刻度的旋钮——薪资底线、外劳税、配额——并且接受了一个义务：每年公开地、带着数字去转动这个旋钮。多数国家把移民参数当条约：一次设定、永远捍卫；新加坡把它当利率：临时设定、持续复审。

就业准证的两级闸门如何运作？

就业准证（EP）——新加坡主要的技术工作签证——如今分两阶段运行，第一级就是锚文所说的价格信号的字面实现。第一级是合格薪资，按行业和年龄锚定本地 PMET（专业人士、经理、高管、技师）工资分布的前三分之一。自 2025 年起，新申请人一般行业月薪至少 5,600 新元、金融业 6,200 新元；2026 年升至 5,800 / 6,400 新元，2027 年起升至 6,000 / 6,600 新元。薪资底线随年龄递增，四十五岁上下的申请人约达 10,700 新元。请细读这个设计：门槛不是议会里捍卫的政治数字，而是本地工资分布的函数。本地专业人士越富裕，闸门自动越紧。外籍招聘被强制保持为本地工资的「互补品」而非「替代品」——这正是锚文所说理性体系必须具备的性质，也恰恰是澳大利亚冻结的打分制与美国 1990 年古董配额双双缺失的性质。

COMPASS 如何把别国交给自由裁量的东西变成分数？

第二级是 2023 年 9 月对新申请生效、2024 年 9 月扩展至续签（2026 年 1 月又有更新）的 COMPASS——互补性评估框架。申请人需在四个基础维度上拿满 40 分，每项计 0、10 或 20 分。C1 薪资项按同行业同年龄段本地 PMET 工资分位计分：低于第 65 分位得零分，第 65 至 90 分位得 10 分，第 90 分位及以上得 20 分。C2 学历项对照顶级院校名单。C3 多元化项：若申请人国籍已占该公司 PMET 员工 25% 以上，得零分；5% 至 25% 得 10 分；低于 5% 得 20 分。C4 本地就业支持项：考察公司本地 PMET 占比相对同业的表现。另有两个加分项：紧缺职业清单（C5，最高 20 分）与战略经济优先（C6，10 分）。月薪不低于 22,500 新元者、跨国公司内部调动者及极短期任职者豁免；PMET 少于 25 人的小公司在 C3、C4 上默认各得 10 分；人力部还提供免费自评工具，让雇主在递交前就能算出分数。这里的概念飞跃很容易被忽略：C3 和 C4 考的是企业，不是个人。德国机会卡只问申请人是否合格；COMPASS 还要追问雇主的招聘结构是否健康。多元化与本地就业两项，把外部性——工资压制、族裔飞地化招聘、技能空心化——重新定价回制造它的那一方。这是需求侧校准，世界上没有任何其他打分制做得如此直白。

外劳税与比例配额如何把承载力约束工程化到工作准证层？

如果说 EP 与 COMPASS 校准的是技能分布的顶端，S 准证与工作准证层就是新加坡把锚文的承载力约束从修辞变成工程的地方。S 准证面向中级技术人员，合格薪资自 2025 年 9 月起为 3,300 新元（金融业 3,800 新元），2027 年升至 3,600 / 4,000 新元，同样随年龄递增。每一名 S 准证持有者都让雇主付出外劳税：自 2025 年 9 月起统一为每月 650 新元（此前基础档为 550 新元）。总量则由比例配额硬封顶：S 准证持有者不得超过公司劳动力的 10%（服务业）或 15%（建筑、制造、海事、加工业）。另有两个细节可见这套系统的工艺水准。自 2026 年 7 月起，本地合格薪资上调至 1,800 新元——本地员工月薪达到此线才算一个完整配额单位，900 至 1,800 新元之间只算半个——防止雇主用名义职位虚增本地分母。旋钮也双向转动：2025 年 6 月，工作准证的非传统来源国名单扩展至不丹、柬埔寨、老挝，恰好在存在部门承载力的地方放松供给。税加配额意味着新加坡对同一流量同时操作价格工具和数量工具——相当于政策上既有刹车、又有方向盘。

可复制的边界在哪里——香港、迪拜、卢森堡能学吗？

诚实要求说清什么无法转移。新加坡的架构预设了大陆经济体不具备的条件：单一劳动力市场加上完全可执行的企业登记体系；行政上极易管控的边境；由国家直接控制的住房供给；以及一个迄今接受三成非居民比例作为繁荣代价的政体。比例配额尤其预设了每家企业都有可界定的「本地劳动力」分母——这在六百万人的转口港可行，在 G7 经济体则是行政噩梦。香港和迪拜采用了片段——薪资门槛、顶尖毕业生快速通道——但都不对企业的多元化或本地就业行为计分，因为两者都不像新加坡那样严密管控企业层面的分母。卢森堡在精神上最接近，实际仍松一个数量级。但真正可转移的工具是实打实的：随本地繁荣自动收紧的工资指数化门槛；企业层面的外部性计分；为流量定价的外劳税；以及像本地合格薪资那样保证配额分母诚实性的完整性规则。复制工具，而非架构。

新加坡是不是 85% 系数在现实中的最近似实现？

按锚文框架打分。筛选精度：A——COMPASS 是唯一个同时对雇佣合同两侧计分的主要体系。总量校准：A——门槛与工资指数挂钩、每年复审、公开带数字调整。承载力整合：A−——外劳税与比例配额是真正的硬约束，国家主导的住房体系意味着约束条件被承认而非被否认；扣分项是政治容忍度仍是未定价的变量，2020 年大选已证明外籍 PMET 焦虑能移动真实选票。但请注意系统的反应：三年内收紧 COMPASS、上调薪资底线。校准过的体系不是从不感受压力的体系，而是压力到来时旋钮仍然转得动的体系。总体而言，新加坡拿到本系列迄今最高分——不是因为它的参数完美，而是因为它是唯一把参数当作需要重新测量的假设、而非需要捍卫的立场的国家。系列三联画的轮廓至此完整：澳大利亚有机器没有闸门，德国有闸门没有执行，新加坡两者兼备——但其规模与治理条件是更大国家无法整体进口的。第 4 篇将转向萦绕在三国头上的警示案例：美国——一个在 1990 年冻结了核心参数、然后用三十年让法院、游说集团与运气代替新加坡用电子表格完成的校准的国家。

常见问题

COMPASS 下就业准证申请人需要多少分？四个基础维度共需 40 分——薪资相对本地同行业同年龄段 PMET 分位、学历、企业内部国籍多元化、企业本地就业支持——每项最高 20 分，另有紧缺职业与战略经济优先加分。月薪不低于 22,500 新元者豁免，雇主可使用人力部免费自评工具预先算分。

为什么新加坡几乎每年上调 EP 薪资门槛？因为门槛锚定的是本地 PMET 工资分布的前三分之一（分行业、分年龄），而非政治妥协。随本地工资上升——一般行业从 2025 年的 5,600 新元升至 2027 年预定的 6,000 新元——闸门自动收紧，使外籍招聘始终是本地劳动力的互补品而非替代品。再校准本身就是体系，而非对体系的修补。

By Dr. Tong Yin (殷彤博士) · Founder, InsightBridge Global LLC · August 2026 · Part 3 of the Precision Immigration Governance series, following the anchor essay, Part 1: Australia, and Part 2: Germany.

Part 1 of this series showed that Australia built the Anglosphere's most refined selection machine and then ran it for two decades with the intake gate essentially uncalibrated. Part 2 showed that Germany, after fifty years of denial, designed an A-minus machine that its own administrative machinery executes at C-plus. Both failures, different as they look, share one root: the parameter was set once and left to drift while the economy moved. Part 3 turns to the one country that refused that failure mode from the beginning. Singapore is, to my knowledge, the only government on earth that runs immigration policy the way a central bank runs monetary policy — as a continuously calibrated operating system rather than a recurring political crisis. It is also the closest real-world implementation of the anchor essay's central prescription: hold the system near its measured capacity, never beyond it, and recalibrate before the gap becomes a grievance.

Why is Singapore the only country that treats immigration as an operating system?

The anchor essay argued that optimal immigration policy should be rules-based, countercyclical, and frequently recalibrated — closer to inflation targeting than to legislation. Singapore is the only major economy that actually built this. The structural reasons are partly unique: a city-state of roughly six million with no interior regions, no federal veto players, a housing supply dominated by the state (the HDB houses roughly four-fifths of residents), and a Ministry of Manpower with a genuine technocratic mandate. Non-residents constitute roughly thirty percent of the total population — a share that would trigger political rupture almost anywhere else, and that in Singapore is managed as an operating variable. But the deeper reason is doctrinal. Since the 1990s, MOM has treated the foreign workforce as a dial with measurable settings — salary floors, levies, quotas — and has accepted the obligation to turn the dial every year, in public, with numbers. Most countries treat immigration parameters as treaties: set them once, defend them forever. Singapore treats them as interest rates: set them provisionally, review them continuously.

How does the two-stage Employment Pass gate work?

The Employment Pass — Singapore's main skilled-work visa — now operates in two stages, and the first is the anchor framework's price signal made literal. Stage 1 is a qualifying salary benchmarked to the top one-third of local PMET (professionals, managers, executives, technicians) wages, differentiated by sector and by age. From 2025, new applicants need at least S$5,600 per month in general sectors and S$6,200 in financial services; the schedule rises to S$5,800 and S$6,400 in 2026 and to S$6,000 and S$6,600 from 2027. The floor scales upward with age, reaching roughly S$10,700 for candidates in their mid-forties. Read that design slowly. The threshold is not a political number defended in parliament; it is a function of the local wage distribution. When local professionals get richer, the gate tightens automatically. Foreign hiring is forced to remain a complement to local wages rather than a substitute for them — which is precisely the property the anchor essay said a rational system must have, and precisely the property Australia's frozen points test and America's 1990-vintage caps both lack.

How does COMPASS score what other countries leave to discretion?

Stage 2, introduced for new applications in September 2023 and extended to renewals in September 2024 (with an update effective January 2026), is COMPASS — the Complementarity Assessment Framework. A candidate needs 40 points across four foundational criteria, each scored 0, 10, or 20. C1 scores salary against the percentile of local PMET wages in the same sector and age band: below the 65th percentile earns zero, the 65th to 90th earns 10, at or above the 90th earns 20. C2 scores qualifications against a list of top-tier institutions. C3 scores diversity: if the candidate's nationality already makes up 25 percent or more of the firm's PMET workforce, the score is zero; between 5 and 25 percent, 10; under 5 percent, 20. C4 scores the firm's support for local employment — its share of local PMETs relative to industry peers. Two bonus criteria add points for shortage-occupation roles (C5, up to 20) and strategic economic priorities (C6, 10). Candidates earning at least S$22,500 per month, intra-corporate transferees, and very short assignments are exempt; small firms with fewer than 25 PMETs receive a default 10 points on C3 and C4; and MOM publishes a free self-assessment tool so employers can compute the score before filing. The conceptual leap here is easy to miss: C3 and C4 score the firm, not the individual. Germany's Opportunity Card asks whether the applicant is qualified; COMPASS additionally asks whether the employer's hiring pattern is healthy. The diversity and local-employment criteria push the externality — wage suppression, enclave hiring, skills-hollowing — back onto the party that creates it. This is demand-side calibration, and no other points system in the world does it as explicitly.

How do levies and ratio caps engineer the carrying-capacity constraint at the work-permit layer?

If EP and COMPASS calibrate the top of the skill distribution, the S Pass and Work Permit tiers are where Singapore converts the anchor essay's carrying-capacity constraint from rhetoric into engineering. The S Pass — for mid-skilled technicians — carries a qualifying salary of S$3,300 from September 2025 (S$3,800 in financial services), rising to S$3,600 and S$4,000 by 2027, again age-scaled. Every S Pass holder costs the employer a levy, unified at S$650 per month from September 2025 (up from S$550 at the basic tier). And the total number is hard-capped by the dependency ratio ceiling: S Pass holders may not exceed 10 percent of a firm's workforce in services, or 15 percent in construction, manufacturing, marine, and process. Two further details show the system's craftsmanship. From July 2026, the Local Qualifying Salary rises to S$1,800 — a local employee counts as one full quota unit only at that wage, and as half a unit between S$900 and S$1,800 — which prevents employers from padding the local denominator with nominal jobs. And the dial turns in both directions: in June 2025, the Non-Traditional Sources list for work permits was expanded to Bhutan, Cambodia, and Laos, loosening supply exactly where sectoral capacity existed. Levy plus ceiling means Singapore operates a price instrument and a quantity instrument simultaneously on the same flow — the policy equivalent of having both a brake and a steering wheel.

Where are the limits of replicability — can Hong Kong, Dubai, or Luxembourg copy this?

Honesty requires stating what does not transfer. Singapore's architecture assumes things a continental economy does not have: a single labor market with a fully enforceable employer registry; a border that is administratively trivial to police; housing supply under direct state control; and a polity that has, so far, accepted a thirty-percent non-resident share as the price of prosperity. The dependency ratio ceiling in particular presumes a definable local-workforce denominator for every firm — feasible in a six-million-person entrepôt, administratively brutal in a G7 economy. Hong Kong and Dubai have adopted fragments — salary thresholds, fast tracks for top graduates — but neither scores employers' diversity or local-employment behavior, because neither polices the firm-level denominator as tightly. Luxembourg comes closest in spirit and is still an order of magnitude looser. The transferable instruments are real, though: wage-indexed thresholds that tighten automatically with local prosperity; firm-level externality scoring; levies that price volume; and integrity rules like the Local Qualifying Salary that keep quota denominators honest. Copy the instruments, not the architecture.

Is Singapore the closest real-world approximation of the 85% coefficient?

Score it against the anchor framework. Selection precision: A — COMPASS is the only major system that scores both sides of the employment contract. Intake calibration: A — thresholds are wage-indexed, reviewed annually, and adjusted in public with numbers. Capacity integration: A-minus — the levy and ratio ceilings are genuine hard constraints, and the state-controlled housing system means the binding constraint is admitted rather than denied; the minus is that political tolerance remains the unpriced variable, as the 2020 general election showed when anxiety over foreign PMETs moved real votes. But note how the system responded: it tightened COMPASS and raised salary floors within three years. A calibrated system is not one that never feels pressure; it is one whose dials still turn when pressure arrives. Overall, Singapore earns the strongest grade in this series so far — not because its parameters are perfect, but because it is the only country that treats its parameters as hypotheses to be re-measured rather than positions to be defended. The series' triptych is now complete in outline. Australia has the machine without the gate. Germany has the gate without the execution. Singapore has both — at a scale and under governability conditions that larger states cannot import wholesale. Part 4 turns to the cautionary case that haunts all three: the United States, which froze its core parameters in 1990 and has spent three decades letting courts, lobbyists, and luck do the calibration that Singapore does by spreadsheet.

Frequently Asked Questions

What score does an Employment Pass candidate need under COMPASS? Forty points across four foundational criteria — salary relative to local sector-age PMET percentiles, qualifications, nationality diversity within the firm, and the firm's support for local employment — each worth up to 20 points, plus bonus points for shortage-occupation roles and strategic economic priorities. Candidates earning at least S$22,500 per month are exempt, and employers can pre-compute the score with MOM's free self-assessment tool.

Why does Singapore raise EP salary thresholds almost every year? Because the threshold is benchmarked to the top one-third of local PMET wages by sector and age, not to a political compromise. As local wages rise — from S$5,600 in 2025 to a scheduled S$6,000 in 2027 for general sectors — the gate tightens automatically, keeping foreign hiring a complement to local labor rather than a substitute. Recalibration is the system, not a repair of it.