⟵ 回喵喵咖啡厅⟵ Back to Meowbot Cafe

去哪国发展 · 十三维评估Where to Develop · A 13-Dimension Evaluation18 个常见目的地国家 · 十三维打分 · 每栏都能一格一格复算18 common destination countries · 13 dimensions scored · every column can be recomputed cell by cell

这套评估把“去哪发展”拆成十三栏:经济基本面、制度质量、生活成本、产业能力、薪资、养老代际、税负、未来压力、教育与研究、安全、移民难度、宜居、吃喝玩乐。

This evaluation breaks where-to-develop into 13 columns: economic fundamentals, institutional quality, cost of living, industrial capability, wages, pensions and generational transfer, tax burden, future pressure, education and research, safety, immigration difficulty, livability, and food and leisure.

数据全部来自公开官方源(World Bank、Transparency International、Numbeo、WIPO、IFR、OECD、IMF、CWUR),每个数字都能追到来源与年份。

All data come from public official sources (World Bank, Transparency International, Numbeo, WIPO, IFR, OECD, IMF, CWUR); every number can be traced to its source and year.

它不给单一排名,只给分布:同一个国家在不同权重下的名次会变,第三节的滑条就是让你自己调。

It does not give a single ranking but a distribution: the same country changes rank under different weights, and the sliders in the third section let you set them yourself.

一、十三栏的精确含义与算法1. What the 13 columns mean, and exactly how they are computed

统一口径:每个指标取 18 国的 5% 分位与 95% 分位作为上下界(截尾极差),z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1);成本型指标(越小越好)取 1 − z。维度分 = 该维各指标 z 的等权平均。这样单个离群值不会把整条尺度拉歪,而且每一格都能用原始值重算。

Common convention: for each indicator the 5th and 95th percentiles across the 18 countries are used as bounds (a winsorised range), z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. This keeps a single outlier from skewing the whole scale, and lets every cell be recomputed from raw values.

D1经济基本面

一个国家的“大盘子”和宏观稳定:人均产出有多高、近几年增长快不快、物价稳不稳、失业多不多。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
人均 GDP(PPP)越大越好 +World Bank54616.18 — 135133.63
GDP 五年增长越大越好 +World Bank-1.48 — 3.03
通胀率越小越好 −World Bank3.1 — 7.92
失业率越小越好 −World Bank2.66 — 7.38
新加坡0.78
美国0.73
瑞士0.68
韩国0.65
澳大利亚0.62
日本0.61
挪威0.59
丹麦0.55
荷兰0.54
加拿大0.54
爱尔兰0.50
新西兰0.49
德国0.42
英国0.35
法国0.34
奥地利0.22
芬兰0.14
瑞典0.11

D2制度质量

政府治理水平与腐败程度:六个治理维度的打分加上腐败感知指数。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
WGI 话语权与问责越大越好 +wb:GOV_WGI_CC.EST1.15 — 2.22
WGI 政府效能越大越好 +wb:GOV_WGI_GE.EST1.27 — 2.07
WGI 政治稳定越大越好 +wb:GOV_WGI_PV.EST-0.02 — 1.29
WGI 监管质量越大越好 +wb:GOV_WGI_RQ.EST1.1 — 1.91
WGI 法治越大越好 +wb:GOV_WGI_RL.EST1.02 — 1.99
WGI 腐败控制越大越好 +wb:GOV_WGI_VA.EST0.71 — 1.81
腐败感知指数 CPI越大越好 +Transparency Intl64.85 — 88.3
丹麦0.93
芬兰0.87
新西兰0.83
瑞士0.80
挪威0.79
新加坡0.70
瑞典0.70
荷兰0.66
澳大利亚0.64
爱尔兰0.61
加拿大0.52
日本0.51
德国0.49
奥地利0.40
英国0.32
韩国0.19
法国0.12
美国0.09

D3生活成本

日常与住房的贵不贵:生活成本指数越高越贵,房价收入比越高越难买房。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
Numbeo 生活成本指数越小越好 −Numbeo55.34 — 93.65
房价收入比越小越好 −Numbeo5.78 — 24.45
加拿大0.88
新西兰0.88
日本0.84
瑞典0.83
美国0.81
芬兰0.81
爱尔兰0.79
法国0.78
英国0.78
德国0.78
澳大利亚0.73
荷兰0.72
奥地利0.68
丹麦0.67
挪威0.57
韩国0.48
瑞士0.35
新加坡0.04

D4产业能力

科技与制造业的产业深度:创新指数排名越靠前越好,每万名工人里的机器人越多越好。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
全球创新指数 GII 排名越小越好 −WIPO GII1.85 — 23.3
每万工人机器人密度越大越好 +IFR285.75 — 917.5
瑞士1.00
韩国0.90
英国0.85
芬兰0.76
新加坡0.75
荷兰0.71
丹麦0.62
法国0.53
瑞典0.52
美国0.47
加拿大0.43
德国0.42
日本0.33
奥地利0.29
爱尔兰0.20
挪威0.11
澳大利亚0.01
新西兰0.00

D5薪资

税前薪资水平,三项等权:平均年薪、中位年薪、前 10% 档(P90),都按购买力平价折算。中位与 P90 由 OECD 十分位收入比经对数正态桥推算。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
平均年薪(PPP)越大越好 +OECD58423.07 — 88038.72
中位年薪(PPP,推算)越大越好 +wage:median50199.81 — 75353.86
前 10% 档 P90(PPP,推算)越大越好 +wage:p9090084.14 — 154759.46
瑞士0.99
美国0.92
荷兰0.73
奥地利0.68
丹麦0.67
德国0.62
挪威0.52
澳大利亚0.49
爱尔兰0.43
加拿大0.36
英国0.31
芬兰0.20
韩国0.15
瑞典0.14
新西兰0.13
法国0.12
日本0.00
新加坡缺

D6养老代际

退休后的替代水平与公共养老负担:净替代率与总替代率越高越好,公共养老支出占比高说明代际负担重。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
净替代率 NPRR100越大越好 +OECD41.7 — 96.94
总替代率 GPRR100越大越好 +OECD31.58 — 74.22
公共养老支出 / GDP越大越好 +OECD3.31 — 13.52
奥地利0.94
荷兰0.76
法国0.70
丹麦0.67
芬兰0.64
瑞典0.55
美国0.53
德国0.48
加拿大0.30
英国0.30
挪威0.30
新西兰0.28
日本0.23
瑞士0.23
爱尔兰0.18
澳大利亚0.14
韩国0.03
新加坡缺

D7税负

打工人和企业被拿走多少:税楔算的是“公司在你工资之前缴掉的那部分也一并算进来”。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
税楔越小越好 −OECD22.58 — 47.09
资本利得税越小越好 −OECD0 — 42.9
企业所得税越小越好 −OECD8.93 — 30.92
瑞士0.99
新加坡0.82
加拿大0.81
新西兰0.57
爱尔兰0.54
美国0.49
日本0.49
英国0.45
韩国0.40
澳大利亚0.39
荷兰0.39
德国0.36
瑞典0.33
挪威0.32
丹麦0.29
芬兰0.28
奥地利0.24
法国0.07

D8未来

未来的人口与财政压力:总和生育率、老龄化与政府债务都是成本项;本维先取三项截尾 z 的平均,再整维乘一次生育率惩罚因子(只在 TFR 小于 1.5 时生效),最后在 18 国内重新归一化。

本维特殊规则D8 分三步:① 生育率以普通 5%/95% 截尾 z 进入本维平均(不再单独吃惩罚);② 本维总分再整维乘一次惩罚因子 g:TFR 大于等于 1.5 时 g = 1(无影响),TFR 小于 1.5 时 g = exp(−2.0 × (1.5 − TFR)/0.5);③ 乘完在 18 国内重新归一化到 [0,1] 区间
怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
总和生育率 TFR分段函数World Bank0.93 — 1.62
65 岁以上人口占比越小越好 −World Bank15.17 — 24.48
政府债务 / GDP(IMF)越小越好 −IMF32.87 — 177.88
爱尔兰1.00
新西兰0.97
澳大利亚0.90
丹麦0.82
美国0.78
英国0.73
瑞典0.64
法国0.64
荷兰0.57
挪威0.54
德国0.37
瑞士0.37
奥地利0.29
加拿大0.21
芬兰0.16
新加坡0.02
日本0.02
韩国0.00

D9教育

好大学的存量与科研产出:大学按名次分档加权,再加人均 SCI 论文与研发强度。

本维特殊规则第 n 档 = 前 2 的 n 次方 × 50 名;该档权重 = 0.8 的 n 次方(等价于 (名次/50) 的 −0.3219 次方)
怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
大学分层分(幂律)越大越好 +CWUR 20243 — 58.62
人均 SCI 论文越大越好 +World Bank88.93 — 257.42
研发强度 / GDP越大越好 +World Bank1.58 — 3.8
美国0.69
瑞士0.60
瑞典0.56
丹麦0.56
韩国0.54
英国0.50
日本0.50
德国0.49
芬兰0.46
奥地利0.39
挪威0.37
澳大利亚0.36
荷兰0.31
法国0.27
加拿大0.26
新加坡0.26
爱尔兰0.15
新西兰0.13

D10安全

治安:凶杀率与犯罪感知指数都是成本项。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
凶杀率越小越好 −World Bank0.21 — 2.55
Numbeo 犯罪指数越小越好 −Numbeo22.29 — 50.19
新加坡1.00
日本0.99
瑞士0.81
韩国0.81
荷兰0.80
丹麦0.78
芬兰0.76
奥地利0.73
挪威0.69
德国0.55
爱尔兰0.42
澳大利亚0.41
瑞典0.34
英国0.34
法国0.26
新西兰0.26
加拿大0.20
美国0.02

D11移民难度

(正规)移民难度:学术/工作/投资三条路中,拿到永居所需的最短期望年限(成本型,越短越好)。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
academic越小越好 −PR years, academic route (cost)1.44 — 8.5
work越小越好 −PR years, work route (cost)1.44 — 10
invest越小越好 −PR years, investor route (cost)2 — 10.5
加拿大1.00
新西兰0.89
澳大利亚0.87
德国0.85
挪威0.81
爱尔兰0.77
英国0.72
芬兰0.69
法国0.56
荷兰0.56
奥地利0.56
韩国0.56
日本0.54
美国0.33
丹麦0.12
瑞士0.01
瑞典缺
新加坡缺

D12宜居

宜居:舒适气温天数(收益)+ 酷热/严寒/雨日/PM2.5/自然灾害/病媒(成本)。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
comfort_days越大越好 +D12 raw (Open-Meteo/WHO/INFORM)61.96 — 138.34
hot30越小越好 −D12 raw (Open-Meteo/WHO/INFORM)0 — 1.33
cold0越小越好 −D12 raw (Open-Meteo/WHO/INFORM)0 — 80.74
rain_days越小越好 −D12 raw (Open-Meteo/WHO/INFORM)96.91 — 184.45
pm25越小越好 −D12 raw (Open-Meteo/WHO/INFORM)4.97 — 12.96
hazard_inform越小越好 −D12 raw (Open-Meteo/WHO/INFORM)1.21 — 2.86
vector_aedes越小越好 −D12 raw (Open-Meteo/WHO/INFORM)0 — 9.83
新西兰0.88
丹麦0.77
英国0.73
澳大利亚0.72
瑞典0.70
奥地利0.69
瑞士0.68
德国0.67
法国0.65
挪威0.64
爱尔兰0.64
荷兰0.63
芬兰0.62
加拿大0.53
日本0.50
新加坡0.47
韩国0.46
美国0.41

D13吃喝玩乐

吃喝玩乐(含自然风光名胜):世界遗产密度、保护地占比、森林覆盖、游客量、餐饮价格(成本)、娱乐餐饮消费占比。

怎么算:每个指标先做 5%/95% 截尾归一化 z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1),成本型取 1 − z;本维得分 = 各指标 z 的等权平均。下表的上界就是实际用到的截尾值,可逐格复算。
指标方向来源5%–95% 上下界
heritage_per10m越大越好 +D13 raw (UNESCO/WB/Numbeo/OECD)1.55 — 15.21
protected_pct越大越好 +D13 raw (UNESCO/WB/Numbeo/OECD)11.31 — 34.28
forest_pct越大越好 +D13 raw (UNESCO/WB/Numbeo/OECD)11.46 — 69.45
arrivals_pc越大越好 +D13 raw (UNESCO/WB/Numbeo/OECD)0.33 — 3.8
restaurant_index越小越好 −D13 raw (UNESCO/WB/Numbeo/OECD)32.5 — 92.12
fun_share越大越好 +D13 raw (UNESCO/WB/Numbeo/OECD)0.13 — 0.21
奥地利0.76
瑞典0.52
法国0.52
韩国0.51
新西兰0.49
日本0.47
德国0.42
芬兰0.41
丹麦0.40
挪威0.38
爱尔兰0.34
新加坡0.33
瑞士0.30
荷兰0.30
加拿大0.30
英国0.29
澳大利亚0.29
美国0.17

D1Economic fundamentals

The overall size and macro stability of a country: how high output per person is, how fast it has been growing, how stable prices are, and how much unemployment there is.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
GDP per capita (PPP)higher is better +World Bank54616.18 — 135133.63
Real GDP growth, 5yhigher is better +World Bank-1.48 — 3.03
Inflation ratelower is better −World Bank3.1 — 7.92
Unemployment ratelower is better −World Bank2.66 — 7.38
Singapore0.78
United States0.73
Switzerland0.68
South Korea0.65
Australia0.62
Japan0.61
Norway0.59
Denmark0.55
Netherlands0.54
Canada0.54
Ireland0.50
New Zealand0.49
Germany0.42
United Kingdom0.35
France0.34
Austria0.22
Finland0.14
Sweden0.11

D2Institutional quality

Government quality and corruption: six governance scores plus the Corruption Perceptions Index.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
WGI Voice and Accountabilityhigher is better +wb:GOV_WGI_CC.EST1.15 — 2.22
WGI Government Effectivenesshigher is better +wb:GOV_WGI_GE.EST1.27 — 2.07
WGI Political Stabilityhigher is better +wb:GOV_WGI_PV.EST-0.02 — 1.29
WGI Regulatory Qualityhigher is better +wb:GOV_WGI_RQ.EST1.1 — 1.91
WGI Rule of Lawhigher is better +wb:GOV_WGI_RL.EST1.02 — 1.99
WGI Control of Corruptionhigher is better +wb:GOV_WGI_VA.EST0.71 — 1.81
Corruption Perceptions Indexhigher is better +Transparency Intl64.85 — 88.3
Denmark0.93
Finland0.87
New Zealand0.83
Switzerland0.80
Norway0.79
Singapore0.70
Sweden0.70
Netherlands0.66
Australia0.64
Ireland0.61
Canada0.52
Japan0.51
Germany0.49
Austria0.40
United Kingdom0.32
South Korea0.19
France0.12
United States0.09

D3Cost of living

How expensive daily life and housing are: a higher cost-of-living index means pricier, and a higher price-to-income ratio means it is harder to buy a home.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Numbeo Cost of Living Indexlower is better −Numbeo55.34 — 93.65
Price-to-income ratiolower is better −Numbeo5.78 — 24.45
Canada0.88
New Zealand0.88
Japan0.84
Sweden0.83
United States0.81
Finland0.81
Ireland0.79
France0.78
United Kingdom0.78
Germany0.78
Australia0.73
Netherlands0.72
Austria0.68
Denmark0.67
Norway0.57
South Korea0.48
Switzerland0.35
Singapore0.04

D4Industrial capability

The depth of the tech and manufacturing base: a better innovation rank and more robots per 10,000 workers are both better.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Global Innovation Index (GII) ranklower is better −WIPO GII1.85 — 23.3
Robots per 10,000 workershigher is better +IFR285.75 — 917.5
Switzerland1.00
South Korea0.90
United Kingdom0.85
Finland0.76
Singapore0.75
Netherlands0.71
Denmark0.62
France0.53
Sweden0.52
United States0.47
Canada0.43
Germany0.42
Japan0.33
Austria0.29
Ireland0.20
Norway0.11
Australia0.01
New Zealand0.00

D5Wages

Pre-tax wages, three items equally weighted: average annual wage, median wage and the top 10 percent band (P90), all in purchasing-power-parity terms. Median and P90 are bridged from OECD decile ratios with a lognormal fit.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Average annual wage (PPP)higher is better +OECD58423.07 — 88038.72
Median annual wage (PPP, estimated)higher is better +wage:median50199.81 — 75353.86
Top-10% P90 wage (PPP, estimated)higher is better +wage:p9090084.14 — 154759.46
Switzerland0.99
United States0.92
Netherlands0.73
Austria0.68
Denmark0.67
Germany0.62
Norway0.52
Australia0.49
Ireland0.43
Canada0.36
United Kingdom0.31
Finland0.20
South Korea0.15
Sweden0.14
New Zealand0.13
France0.12
Japan0.00
Singaporemissing

D6Pensions and generational transfer

Post-retirement replacement and the public pension burden: higher net and gross replacement rates are better, and a higher public pension share of GDP means a heavier generational burden.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Net pension replacement ratehigher is better +OECD41.7 — 96.94
Gross pension replacement ratehigher is better +OECD31.58 — 74.22
Public pension spending / GDPhigher is better +OECD3.31 — 13.52
Austria0.94
Netherlands0.76
France0.70
Denmark0.67
Finland0.64
Sweden0.55
United States0.53
Germany0.48
Canada0.30
United Kingdom0.30
Norway0.30
New Zealand0.28
Japan0.23
Switzerland0.23
Ireland0.18
Australia0.14
South Korea0.03
Singaporemissing

D7Tax burden

How much is taken from workers and firms: the tax wedge also counts the part the employer pays before your wage.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Tax wedgelower is better −OECD22.58 — 47.09
Capital gains tax, top ratelower is better −OECD0 — 42.9
Statutory corporate income taxlower is better −OECD8.93 — 30.92
Switzerland0.99
Singapore0.82
Canada0.81
New Zealand0.57
Ireland0.54
United States0.49
Japan0.49
United Kingdom0.45
South Korea0.40
Australia0.39
Netherlands0.39
Germany0.36
Sweden0.33
Norway0.32
Denmark0.29
Finland0.28
Austria0.24
France0.07

D8Future pressure

Future demographic and fiscal pressure: fertility, ageing and government debt are all cost items. This dimension first takes the winsorised z average of the three, then multiplies the whole dimension once by a fertility penalty factor (only when TFR is below 1.5), and finally renormalises across the 18 countries.

Special rule for this dimensionD8 has three steps. (1) Fertility enters the dimension average as an ordinary 5%/95% winsorised z (no separate penalty). (2) The dimension total is then multiplied once, for the whole dimension, by a penalty factor g: g = 1 when TFR is at least 1.5 (no effect), and g = exp(−2.0 × (1.5 − TFR)/0.5) when TFR is below 1.5. (3) After multiplying, results are renormalised across the 18 countries onto [0,1].
How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Total fertility rate (TFR)piecewiseWorld Bank0.93 — 1.62
Population aged 65+lower is better −World Bank15.17 — 24.48
Government debt / GDP (IMF)lower is better −IMF32.87 — 177.88
Ireland1.00
New Zealand0.97
Australia0.90
Denmark0.82
United States0.78
United Kingdom0.73
Sweden0.64
France0.64
Netherlands0.57
Norway0.54
Germany0.37
Switzerland0.37
Austria0.29
Canada0.21
Finland0.16
Singapore0.02
Japan0.02
South Korea0.00

D9Education and research

The stock of good universities and research output: universities weighted by rank tier, plus SCI papers per capita and R and D intensity.

Special rule for this dimensionTier n = the top 2 to the power n times 50 universities; the weight of that tier = 0.8 to the power n (equivalently, rank divided by 50 raised to the power −0.3219).
How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
University tier score (power law)higher is better +CWUR 20243 — 58.62
SCI papers per capitahigher is better +World Bank88.93 — 257.42
R and D intensity / GDPhigher is better +World Bank1.58 — 3.8
United States0.69
Switzerland0.60
Sweden0.56
Denmark0.56
South Korea0.54
United Kingdom0.50
Japan0.50
Germany0.49
Finland0.46
Austria0.39
Norway0.37
Australia0.36
Netherlands0.31
France0.27
Canada0.26
Singapore0.26
Ireland0.15
New Zealand0.13

D10Safety

Public safety: the homicide rate and the crime perception index are both cost items.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
Homicide ratelower is better −World Bank0.21 — 2.55
Numbeo Crime Indexlower is better −Numbeo22.29 — 50.19
Singapore1.00
Japan0.99
Switzerland0.81
South Korea0.81
Netherlands0.80
Denmark0.78
Finland0.76
Austria0.73
Norway0.69
Germany0.55
Ireland0.42
Australia0.41
Sweden0.34
United Kingdom0.34
France0.26
New Zealand0.26
Canada0.20
United States0.02

D11Immigration difficulty

(Regular) immigration difficulty: the shortest expected number of years to permanent residence across the academic, work and investment routes (a cost item, so shorter is better).

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
academiclower is better −PR years, academic route (cost)1.44 — 8.5
worklower is better −PR years, work route (cost)1.44 — 10
investlower is better −PR years, investor route (cost)2 — 10.5
Canada1.00
New Zealand0.89
Australia0.87
Germany0.85
Norway0.81
Ireland0.77
United Kingdom0.72
Finland0.69
France0.56
Netherlands0.56
Austria0.56
South Korea0.56
Japan0.54
United States0.33
Denmark0.12
Switzerland0.01
Swedenmissing
Singaporemissing

D12Livability

Livability: comfortable-temperature days (a benefit) plus heat, cold, rainy days, PM2.5, natural hazards and disease vectors (costs).

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
comfort_dayshigher is better +D12 raw (Open-Meteo/WHO/INFORM)61.96 — 138.34
hot30lower is better −D12 raw (Open-Meteo/WHO/INFORM)0 — 1.33
cold0lower is better −D12 raw (Open-Meteo/WHO/INFORM)0 — 80.74
rain_dayslower is better −D12 raw (Open-Meteo/WHO/INFORM)96.91 — 184.45
pm25lower is better −D12 raw (Open-Meteo/WHO/INFORM)4.97 — 12.96
hazard_informlower is better −D12 raw (Open-Meteo/WHO/INFORM)1.21 — 2.86
vector_aedeslower is better −D12 raw (Open-Meteo/WHO/INFORM)0 — 9.83
New Zealand0.88
Denmark0.77
United Kingdom0.73
Australia0.72
Sweden0.70
Austria0.69
Switzerland0.68
Germany0.67
France0.65
Norway0.64
Ireland0.64
Netherlands0.63
Finland0.62
Canada0.53
Japan0.50
Singapore0.47
South Korea0.46
United States0.41

D13Food and leisure

Food, drink and leisure (including natural scenery): World Heritage density, protected-area share, forest cover, tourist volume, dining prices (a cost) and the share of spending on entertainment and dining.

How it is computed: every indicator is first winsorised and normalised, z = clip((x − q0.05)/(q0.95 − q0.05), 0, 1); cost-type indicators (lower is better) are flipped to 1 − z. A dimension score is the equally weighted mean of its indicator z values. The bounds in the table below are the actual winsorisation values used, so every cell can be recomputed from raw data.
IndicatorDirectionSource5%–95% bounds
heritage_per10mhigher is better +D13 raw (UNESCO/WB/Numbeo/OECD)1.55 — 15.21
protected_pcthigher is better +D13 raw (UNESCO/WB/Numbeo/OECD)11.31 — 34.28
forest_pcthigher is better +D13 raw (UNESCO/WB/Numbeo/OECD)11.46 — 69.45
arrivals_pchigher is better +D13 raw (UNESCO/WB/Numbeo/OECD)0.33 — 3.8
restaurant_indexlower is better −D13 raw (UNESCO/WB/Numbeo/OECD)32.5 — 92.12
fun_sharehigher is better +D13 raw (UNESCO/WB/Numbeo/OECD)0.13 — 0.21
Austria0.76
Sweden0.52
France0.52
South Korea0.51
New Zealand0.49
Japan0.47
Germany0.42
Finland0.41
Denmark0.40
Norway0.38
Ireland0.34
Singapore0.33
Switzerland0.30
Netherlands0.30
Canada0.30
United Kingdom0.29
Australia0.29
United States0.17

二、十三维雷达图2. The 13-dimension radar charts

第一张是六个代表国家叠在一张雷达上对比;第二张是 18 国每个一张小雷达,一眼看出各自的长板与短板。“制度阴间处”是观察项、不计分,不出现在雷达里。

The first chart overlays six representative countries on one radar for comparison; the second gives each of the 18 countries a small radar, so strengths and weak spots show at a glance. The institutional-hellscape item is an observation only, not scored, and does not appear in the radars.

图 4 十三维雷达图(6 国叠加) 十三轴 = 十三个维度,半径 1 = 该维满分。叠 6 国看形状:越接近正圆越均衡,越尖越偏科。 经济基本面 制度质量 生活成本 产业能力 薪资 养老代际 税负 未来 教育 安全 移民难度 宜居 吃喝玩乐 SGP CHE NLD DNK USA JPN 新加坡缺「薪资」「养老」两轴:多边形在那两处直接跨过,属数据缺失而非短板。 瑞士是尖星形(薪资、税负戳得很远,生活成本缩到圆心);美国的安全轴、日本的经济轴都缩到圆心。
图 5 十八国小倍数雷达(统一 0–1 刻度) 每国一张十三轴雷达、半径固定,可横向比形状:越圆越均衡,越尖越偏科。十三根轴自正上方起顺时针依次是:经济基本面、制度质量、生活成本、 产业能力、薪资、养老与代际、税负、未来压力、教育与研究、安全、移民难度、宜居、吃喝玩乐(图中用两字简称)。 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 SGP 缺 薪资/养老 两轴 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 CHE 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 NLD 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 DNK 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 IRL 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 FIN 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 NZL 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 NOR 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 SWE 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 GBR 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 DEU 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 AUT 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 USA 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 CAN 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 AUS 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 KOR 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 JPN 经济 制度 成本 产业 薪资 养老 税负 未来 教育 安全 移民 宜居 玩乐 FRA 数据:data/derivation.json(5%/95% 截尾后归一化,方向统一为 1 = 好)
Fig. 4 13-dimension radar (6 countries overlaid) 13 axes = 13 dimensions; radius 1 = full marks there. Six overlays: rounder shape = more balanced, spikier = more lopsided. Economy Institutions Cost of living Industry Wages Pensions Taxes Future Education Safety Immigration Livability Leisure SGP CHE NLD DNK USA JPN Singapore lacks the Wages and Pensions axes: the polygon skips them. That is missing data, not a weak point. Switzerland is a spiky star (wages and taxes stick out, cost of living shrinks to the centre); the US safety axis and the Japan economy axis both shrink to the centre.
Fig. 5 Small-multiple radars, 18 countries (fixed 0-1 scale) One 13-axis radar per country, fixed radius, so shapes compare side by side: rounder = more balanced. The 13 axes, clockwise from top: economy, institutions, cost of living, industry, wages, pensions, taxes, future pressure, education, safety, immigration, livability, food and leisure (two-word labels in charts). Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure SGP missing Wages/Pensions Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure CHE Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure NLD Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure DNK Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure IRL Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure FIN Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure NZL Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure NOR Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure SWE Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure GBR Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure DEU Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure AUT Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure USA Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure CAN Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure AUS Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure KOR Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure JPN Economy Institut. Cost Industry Wages Pension Taxes Future Edu. Safety Visa Living Leisure FRA Data: data/derivation.json (5%/95% winsorised, then normalised; 1 = good)

三、自己调权重,看名次怎么变3. Set your own weights and watch the ranking change

十三根滑条 = 十三栏的权重。第一次打开时全是 0:把自己在意的那几栏拉起来,名次就出来了。你拖动后的状态会记在这台设备的浏览器里,下次打开还是上次的样子(换设备或清缓存则不保留)。

13 sliders = the weights of the 13 columns. On first open they are all 0: pull up the columns you care about and the ranking appears. Your setting is remembered in this device browser, so next time it looks the same (changing device or clearing the cache loses it).

最高 2 倍、最低 0(0 = 完全不管这一栏)。名次旁边的 ▲▼ 是相对报告基准(十三栏等权)的升降。

Maximum 2x, minimum 0 (0 = ignore this column entirely). The ▲▼ next to a rank is the change relative to the report baseline (equal weights over the 13 columns).

缺维国家:瑞典 缺 移民;新加坡 缺 薪资、养老、移民。缺维时按它可用的维度加权,所以拖它缺的那几栏时它不动。Countries missing data: Sweden missing Immigration; Singapore missing Wages, Pensions, Immigration. When a dimension is missing, a country is weighted over the dimensions it has, so pulling the sliders it lacks does not move it.