一个找不到出处的数字：论奢华酒店可行性研究中的需求依据

引言

过去两年，从北欧到中东，从东南亚到东欧，全球高端与奢华酒店的开发管线持续扩张。在为这些项目背书的可行性研究、投资备忘录和品牌路演材料中，一个数字被反复引用：全球高净值人群增长了 15.5%。

这个数字被用来支撑一个推论：既然富人变多了，奢华酒店的需求就会增长，因此现在是扩张的时候。

本文不讨论任何机构的动机，也不指控任何人。本文只提三个纯技术性的问题：

第一，这个数字的一手出处在哪里？ 第二，即使它准确，财富存量的增长能否推导出住宿需求的增长？ 第三，即使总需求增长，它能否转化为某一个具体项目的客人？

这三个问题都属于统计方法与基础经济学的范畴，答案与任何人的商业立场无关。

一、这个数字的一手出处

我尝试对"全球高净值人群增长 15.5%"做溯源，逐页核查了目前发布高净值人口统计的主要机构的原始报告与新闻稿，包括 Capgemini World Wealth Report、UBS Global Wealth Report、Altrata World Ultra Wealth Report、Knight Frank The Wealth Report、Henley & Partners / New World Wealth，以及旅游业研究机构 Skift 与 Skift Research 的奢华旅游相关报告。

我没有在其中任何一家的公开材料中找到这一数字。

需要说明清楚的是，这不等于该数字一定是错的。它可能来自我未能检索到的付费报告、内部演示或区域性研究。但对于一个正在被用于支撑数十亿美元资本配置的关键参数而言，"检索不到一手出处"本身就是一个应当被正视的问题。任何在可行性研究中引用它的一方，都应当能够出示它的原始来源、定义门槛、测量年份与测量方法。这是引用者的举证责任，不是质疑者的。

顺带澄清两处在行业讨论中常见的归属混淆。Skift 是旅游业的 B2B 媒体与研究公司，并非信用评级机构；其奢华旅游内容中出现的财富数字，是转引 Knight Frank 等第三方，而非自有测算。Moody's Analytics 曾为《华尔街日报》做过一项广受引用的分析——收入前 10% 的美国家庭贡献了接近一半的消费支出——但该分析的口径是美国国内消费支出、按家庭年收入约 25 万美元分组，既不是全球旅行支出，也不是按净资产划分的高净值人群。将这两者混用，会在推理链的第一步就引入口径错误。

二、能找到出处的数字之间，彼此差多少

比"找不到出处"更值得关注的，是那些能找到出处的数字之间的离散程度。

2023 年，同一个净资产门槛（3,000 万美元以上），同一个"超高净值人群"措辞：

两者相差 200,289 人，即 47%；增速相差近一倍。

2025 年的"百万富翁"人口：

同一年，"富裕人群增长率"从 +1.5% 到 +14.4%，取决于选择哪家机构、哪个门槛。

这些差异中有相当一部分是可解释的，而且是各机构自己写明的。Capgemini 将高净值人群定义为可投资资产在 100 万美元以上者，并明确排除自住主宅；UBS 的净财富定义则包含自住房产并扣除债务。两个定义衡量的本就不是同一群人。

但这恰恰是问题所在：一个不注明定义门槛、资产口径与测量年份的"全球高净值人群增长 X%"，在技术意义上是没有定义的。它无法被复核，无法被比较，也无法被用于任何需要精度的测算。

三、这些数字是怎么产生的

要理解为什么会出现 47% 的差距，需要看这些数字的生成方式。各机构在其公开方法论中的自述如下。

Capgemini 采用两阶段专有模型：先由 IMF 与世界银行的国民账户储蓄数据逐年累积得到总财富存量，再按全球股指调整为市值口径；然后使用收入分布数据，通过"财富—收入关系公式"推导财富分布，覆盖 71 个市场。

Altrata / Wealth-X 采用 Wealth and Investable Assets Model：先以世界银行、IMF、OECD 及各国统计机构的数据做计量估计，得出私人财富总量；再基于其专有数据库构建 Lorenz 曲线分配到人群。

UBS 表述为"a model employing macroeconomic variables"，覆盖 56 个市场，含自住房产、扣除负债，以期末汇率折算美元。

Henley & Partners / New World Wealth 使用国别基准模型，输入包括家庭收入统计、股市市值、内部数据库、税务代理变量与高端住宅价格校验。

换言之，这些高净值人口数字都不是计数结果，而是模型输出——由宏观总量反推、再用某种分布假设分配到人头。这本身不是缺陷；在缺乏全球财富普查的现实下，这是唯一可行的做法。

真正需要指出的是随之而来的四项披露状况，均出自各机构自己的公开文件：

上述五家机构中，没有一家披露高净值人口估计的置信区间或误差范围。

没有一家的结果经过独立第三方审计或同行评议。（UBS 报告中 PwC Switzerland 的角色被表述为数据支持，而非对结果出具审计意见。）

Knight Frank 在其 2026 年版中写明"完整方法论可应要求提供"，即不在报告中公开；其数据来源注明包含 Forbes；其出版免责声明写明该报告"is not definitive and it is not to be relied upon in any way"。该机构亦自述其模型是动态的，历史数字"may not be identical to… previous editions"——这意味着其人口序列不宜作为跨年度可比的时间序列使用。

UBS 明确指出"it is primarily changes in foreign exchange rates that alter the relative performance of different economies' wealth"，并举例 2025 年欧元兑美元升值约 9%。但报告未给出"剔除汇率因素后的百万富翁增量"这一分解。对一个以美元门槛划线的人口计数而言，汇率变动本身就会改变跨线人数，因此这项分解的缺失具有实质影响。

值得一提的是，Henley & Partners / New World Wealth 在其方法论中明确写道：其结果应被视为"modeled estimates… illustrative indicators of broad trends… rather than as precise counts"（模型估计……宽泛趋势的示意性指标，而非精确计数）。这句自我限定说得非常准确，也适用于本节讨论的全部数字。问题不在于机构没有说清楚，而在于使用者在把这些"示意性趋势指标"当作精确参数放进财务模型时，丢掉了这句限定。

这类方法学困难在学术界早有系统研究。财富调查对最顶层人群的覆盖不足与差异性无回答已被反复证明（Vermeulen, ECB, 2014；Bach, Thiemann & Zucco, DIW Discussion Paper 1717, 2018）；以 Pareto 或 Lorenz 型分布外推顶层人数时，尾部起点参数的选择会显著改变最终结果（Eckerstorfer et al., Review of Income and Wealth, 2015）；以富豪榜作为顶层数据源的准确性问题亦有专门研究（Wildauer & Heck, GPERC WP92；Raub, Johnson & Newcomb, IRS SOI, 2010）。这些文献批评的是方法的固有难度，而非任何机构的诚信。但它们共同说明一件事：顶层财富人口的估计存在不可忽略的模型不确定性，而这一不确定性目前没有被量化披露。

需要严格说明的是：这些学术文献所证明的偏差方向是低估顶层财富，而非高估。因此从公开信息出发，无法判断商业模型的净偏差方向。可以指出的是不确定性未被披露，不能主张数字被夸大。

四、即使数字准确：存量与流量是两回事

现在做一个让步性假设：假设"高净值人群增长 15.5%"完全准确。它能否推导出奢华酒店需求的增长？

这是一个纯粹的经济学问题，答案取决于三个转化环节，每一环都不是自动成立的。

第一，存量不等于流量。 高净值人口是一个存量概念（某时点上净资产超过某门槛的人数）；酒店需求是一个流量概念（某时段内售出的间夜数）。从存量推导流量，需要一个转化系数——人均年住宿夜数。这个系数必须被独立测量，不能被假定。

第二，纸面财富不等于购买力。 各机构的财富总量口径包含股票市值与房产估值。当资产价格上涨时，跨过美元门槛的人数增加，但这部分增量是未实现的估值变动，不必然转化为当期可支配支出。这一点在现有实证中有直接证据：一项检验财富效应对出境旅游需求影响的研究（韩国，1989 Q1—2009 Q4，N=83，Prais–Winsten FGLS 与 Newey–West OLS 双模型）发现，家庭收入的弹性为 1.50（t=16.46，p<0.01）、住房财富为 0.39（t=2.56，p<0.05），而股市财富的系数为 −0.028（t=−0.60），统计上不显著。作者的结论是股市财富效应假设在两个模型中均被拒绝。

这一发现之所以重要，是因为财富报告中的高净值人口增长，主要正是由股票市值变动驱动的——Capgemini 明确说明其总财富按全球股指调整为市值口径。也就是说，驱动"富人变多"的那个变量，恰恰是现有实证中对旅游需求不显著的那个变量。

第三，口径必须可乘。 假设我们接受"高净值人数 × 人均年间夜数 = 市场规模"这一算式。问题在于，目前所有可获得的富裕人群年均住宿夜数调研——例如 Luxury Institute 的七国调研（年均约 19 个酒店间夜，其中奢华酒店约 11 夜）——分组变量都是家庭收入，没有一项是按净资产分组的。而所有高净值人口统计的分组变量都是净资产。两个乘数来自互不重叠的人群定义，相乘在口径上不成立。

还有一个更基本的约束：酒店库存是不可储存的（perishable），且个体消费存在物理上限。一个人一晚只能占用一间客房。财富总量可以无上限增长，一个人的年度住宿夜数不能。因此，财富总额的增长率与间夜需求的增长率之间,不存在恒等关系。

最后需要如实说明：我没有找到任何公开研究检验过高净值人口数量与奢华酒店间夜需求之间的相关性。 这一因果链在行业报告中被反复叙述——例如有大型顾问机构以百万富翁人口 10 年复合增长率 5.9%、亿万富翁 5.2%、全球财富 9.6% 对比奢华酒店供给复合增长率 2.3%，据此论证需求"exceeds supply growth"——但这些叙述中不含相关系数、弹性估计或显著性检验。

这里必须区分两句话："该关联从未被公开数据检验过"是准确的；"该关联已被证伪"则不成立。 前者是事实陈述，后者超出证据范围。我主张的是前者。而对于一项需要 20 年以上回收期的重资产投资来说，"从未被检验"已经足以构成审慎理由。

五、即使总需求增长：它为什么会到你这里来

假设前四节的所有疑虑都不成立——数字准确，存量顺利转化为流量，全球奢华住宿需求确实在增长。对一个具体项目而言，这仍然不足以构成投资依据。

这是基础微观经济学中最容易在宏观叙事里被跳过的一步：行业总需求（industry demand）与企业面对的需求（firm-level demand）是两条不同的曲线。

一位超高净值人士今年多住了 5 个奢华酒店间夜，这 5 夜会落在全球数千家奢华酒店中的某几家。它落在哪里，取决于目的地选择、季节、航线可达性、品牌关系、既有忠诚度、同伴决策，以及最关键的——同一目的地内有多少间同档次客房在同时争夺这 5 夜。

因此，对任何一个具体项目，正确的问题不是"全球富人是否变多了"，而是一组层层收窄的问题：

目的地捕获率：本目的地在全球顶奢间夜中的份额是多少？过去五年这一份额是上升还是下降？

本地供给变动：未来 36 个月，同一目的地内同档次（可比 ADR 区间）将新增多少间客房？这是分母。

本项目捕获率：在新增供给全部入市后，本项目能取得多少市场份额？依据是什么？

需求的目的地专属性：该目的地是否具备不可替代的到访理由？如果没有，客人为何不去成本更低、可达性更好的替代目的地？

季节性集中度：全年需求集中在多少周内？其余时段的固定成本由什么覆盖？

这五个问题的答案都必须来自间夜口径的实测数据，而不能由任何全球财富总量数字推导出来。一个宏观数字无论多准确，都不含有关于目的地竞争结构的任何信息。

供求原理在这里的应用其实非常朴素：价格与入住率由某一具体市场在某一具体时段内的供给与需求共同决定，而不由全球财富存量决定。 当一个目的地的同档次客房供给增速持续高于其实际售出间夜增速时，无论全球有多少位新增富豪，该目的地的入住率与实际房价都会承压。这不是预测，而是恒等式。

这里可以给出一个正在发生的、间夜口径的观察。马尔代夫是全球顶奢住宿最成熟的目的地之一。2025 年，其床位供给增长 4.3%，而售出床位夜增长 2.4%；进入 2026 年 3 月，在同比新增 4,818 张床位的同时，度假村床位夜同比下降 12%，整体入住率从 65% 降至 57%；马尔代夫金融管理局的序列显示，度假村入住率在 2026 年 6 月为 50.1%。

我在此不对该目的地的长期前景做任何判断——单一目的地的短期波动可能有多种原因，且不能外推至全球。我引用它，仅因为它是少数几个以间夜为口径公开披露供需两侧数据的市场之一，因而可以用来说明一件事：在供给增速持续高于间夜增速的市场里，入住率的走向与全球财富统计无关。 这正是一个投资人真正需要看到的数据形态。

至于近年被广泛讨论的中东超大型文旅项目——例如红海项目二期工程的重新评估——我认为不应被简单地当作反面教材。主权财富基金主导的项目有其国家战略目标，其成本收益框架与私人资本根本不同，其经验不能直接套用到任何一个由私人或家族资本承担全部下行风险的项目上。恰恰相反：正因为主权项目可以承受纯商业口径下无法承受的亏损，其规划逻辑对私人投资者的参考价值是有限的。模仿其规模而不具备其风险承受结构，是一种类比错误。

六、可行性研究应当满足的数据标准

以上讨论指向一组具体的、可操作的要求。这些要求不针对任何机构，而是对所有奢华酒店可行性研究的通用建议。

关于每一个被引用的宏观数字，应当要求提供：

一手出处——发布机构、报告名称、版本年份、页码。转引不能作为出处。

精确定义——净资产还是可投资资产？门槛多少？是否含自住房产？是否扣除负债？以何种货币、何时汇率计价？

测量方法——是计数还是模型估计？若为模型，使用了何种分布假设？

不确定性——置信区间或误差范围。若未披露，应在文件中明确注明"该数字未提供不确定性估计"。

可比性声明——该序列是否可跨年度比较？发布方是否对历史数字做过修订？

关于需求论证，应当要求：

需求以间夜为口径（room-nights / bednights sold），而非财富存量、人口存量、会员数量或未分段位的管线总量。

供需两侧同口径对照——同一目的地、同一 ADR 区间、同一时间窗内的新增供给与实际售出间夜。

捕获率论证明示假设——项目预计取得的市场份额及其依据，须作为可被检验的假设单独列出，而非隐含在总量推导中。

区分行业需求与企业需求——总需求增长不构成本项目需求增长的证明。

区分"未被验证"与"已被证伪"——对于尚无数据支持的因果链，应如实标注为未经检验，并相应提高安全边际，而不是既不标注也不检验。

结语

本文没有主张任何一个数字是伪造的，也没有指控任何机构存在不当行为。事实上，在缺乏全球财富普查的现实条件下，用宏观总量加分布假设去估计高净值人口，是一项技术上极其困难的工作，各机构在方法上的努力值得尊重。其中一家在自己的方法论中明确写下这些结果是"宽泛趋势的示意性指标，而非精确计数"——这是一种应当被赞赏的坦率。

问题出在传递链条上。一个被诚实地标注为"示意性趋势指标"的模型输出，经过若干次转引之后，会失去它的定义、它的口径、它的不确定性说明，最终以一个孤立的百分比出现在投资备忘录的第一页，并被当作精确参数用于测算一项 20 年期的重资产投资。

这个转化过程中没有人撒谎，但结论已经不再可靠。

因此我想提出的建议是克制而具体的：在奢华酒店开发决策中，请把宏观财富数据放回它应有的位置——它是背景，不是依据。 真正能够支撑投资决策的，是目的地层面、间夜口径、供需两侧同时可得的实测数据；是明示的捕获率假设；是对"这些客人为什么会来这里、为什么会住这一家"的具体回答。

数据的科学性不是学术上的洁癖，它是资本的安全边际。当一个数字连一手出处都无法出示时，以它为基础做出的任何测算，其精度都不可能高于这个数字本身。

参考来源

Altrata, World Ultra Wealth Report 2026：https://altrata.com/wp-content/uploads/2026/06/Altrata_World-Ultra-Wealth-Report-2026_FINAL.pdf

Altrata, World Ultra Wealth Report 2024：https://altrata.com/reports/world-ultra-wealth-report-2024

Capgemini, World Wealth Report 2026：https://www.capgemini.com/de-de/wp-content/uploads/sites/8/2026/06/2026-05-26-STUDIE-World-Wealth-Report-2026-1.pdf

Capgemini, World Wealth Report 研究库页面：https://www.capgemini.com/insights/research-library/world-wealth-report/

UBS, Global Wealth Report 2026：https://www.ubs.com/content/dam/assets/wm/static/gwr/global-wealth-report-en-2026.pdf

Knight Frank, The Wealth Report 2024：https://www.knightfrank.com/site-assets/research/reports/the-wealth-report/previous-editions/the-wealth-report-2024.pdf

Knight Frank, The Wealth Report 2025：https://apac.knightfrank.com/hubfs/Research Reports/Residential/Report PDFs/Knight Frank_The Wealth Report 2025.pdf

Knight Frank, The Wealth Report 2026：https://www.knightfrank.fr/fichiers/publications2020/file//146815-the-wealth-report-2026-6a2298f435832632958583.pdf

Henley & Partners / New World Wealth, Methodology：https://www.henleyglobal.com/publications/africa-wealth-report-2025/methodology

Skift, Media Resources：https://skift.com/media-resources/

Skift Research, A Deep Dive Into Luxury Hotels：https://research.skift.com/reports/a-deep-dive-into-luxury-hotels/

Marketplace, "Higher-income Americans drive bigger share of consumer spending"（Moody's Analytics for WSJ）：https://www.marketplace.org/story/2025/02/24/higher-income-americans-drive-bigger-share-of-consumer-spending

Federal Reserve Bank of Minneapolis, "Have US consumers gone K-shaped? A review of the data"：https://www.minneapolisfed.org/article/2026/have-us-consumers-gone-k-shaped-a-review-of-the-data

JLL, 2026 Global Hotel Investment Outlook：https://www.jll.com/content/dam/jllcom/en/global/documents/reports/research-reports/26-insights-global-hotel-investment.pdf

Vermeulen, P. (2014), "How fat is the top tail of the wealth distribution?", ECB：http://www.piketty.pse.ens.fr/files/Vermeulen2014.pdf

Bach, Thiemann & Zucco (2018), "Looking for the Missing Rich", DIW Discussion Paper 1717：https://www.diw.de/documents/publikationen/73/diw_01.c.575768.de/dp1717.pdf

Eckerstorfer et al. (2015), "Correcting for the Missing Rich", Review of Income and Wealth：https://jakob-kapeller.org/images/pubs/2015-Eckerstorferetal-ROIW.pdf

Davies, Sandström, Shorrocks & Wolff (2011), "The Level and Distribution of Global Household Wealth", The Economic Journal：http://piketty.pse.ens.fr/files/DaviesEtal11.pdf

Wildauer & Heck, "Was Pareto right?", GPERC WP92：https://gala.gre.ac.uk/id/eprint/38597/13/38597 WILDAUER_Was_Pareto_right_Is_the_distribution_of wealth_thick_tailed_(REVISED)_2023.pdf

Raub, Johnson & Newcomb (2010), IRS Statistics of Income：http://piketty.pse.ens.fr/files/RaubJohnsonNewcomb2010.pdf

Alvaredo, Berman & Morelli (2024), "Evidence from the Dead", IZA DP 17389：https://docs.iza.org/dp17389.pdf

"Wealth Effect and Demand for Outbound Tourism", UMass ScholarWorks：https://scholarworks.umass.edu/bitstreams/74d6d59b-61fd-4b06-af1c-c105ef105b7b/download

Luxury Institute, Global Hotels LBSI 调研发布：https://www.einpresswire.com/article/299666196/luxury-institute-survey-provides-country-by-country-rankings-of-global-hotel-brands-by-affluent-travelers-from-the-world-s-richest-countries

Visit Maldives, Quarterly Insights Q1 2026：https://corporate.visitmaldives.com/news/quarterly-insights-q1-2026-tourism-performance-and-demand-outlook/

Corporate Maldives, 本地岛屿旅游与到访量报道：https://corporatemaldives.com/local-island-tourism-holds-ground-amid-tourist-arrival-slump/

Maldives Monetary Authority, 度假村入住率序列：https://database.mma.gov.mv/viya/series/219

Introduction

Over the past two years, luxury and ultra-luxury development pipelines have expanded across the Nordics, the Middle East, Southeast Asia and Central and Eastern Europe. In the feasibility studies, investment memoranda and brand presentations supporting these projects, one figure recurs with remarkable consistency: global high-net-worth individuals have grown by 15.5%.

The figure is used to support an inference: more wealthy people exist, therefore demand for luxury hotel accommodation will grow, therefore now is the time to build.

This article makes no claim about anyone's motives and levels no accusation at any organisation. It asks three purely technical questions:

What is the primary source of this figure?

Even if the figure is accurate, does growth in a wealth stock imply growth in accommodation demand?

Even if aggregate demand grows, does it convert into guests at any particular property?

All three belong to statistical methodology and elementary economics. Their answers are independent of anyone's commercial position.

1. The primary source

I attempted to trace the "15.5% growth in global HNWIs" claim, reviewing the original reports and press releases of the principal organisations that publish high-net-worth population estimates: Capgemini's World Wealth Report, UBS's Global Wealth Report, Altrata's World Ultra Wealth Report, Knight Frank's The Wealth Report, Henley & Partners / New World Wealth, together with the luxury-travel research published by Skift and Skift Research.

I did not find this figure in the public materials of any of them.

To be clear about what this does and does not establish: it does not prove the figure is wrong. It may originate in a paywalled report, an internal presentation, or a regional study I was unable to retrieve. But for a parameter now being used to underwrite billions of dollars of capital allocation, the inability to locate a primary source is itself a matter that deserves attention. Any party citing it in a feasibility study should be able to produce the originating publication, the definitional threshold, the measurement year and the measurement method. That burden falls on the party citing the number, not on the party asking about it.

Two attribution errors common in industry discussion are worth correcting here, without implying fault on anyone's part. Skift is a B2B travel media and research company, not a credit rating agency; the wealth figures appearing in its luxury-travel coverage are cited from third parties such as Knight Frank rather than produced in-house. Moody's Analytics did produce a widely quoted analysis for The Wall Street Journal — that the top 10% of US households by income now account for nearly half of all consumer spending — but that analysis measures US domestic consumer spending, segmented by household income of roughly $250,000 or more. It is neither a global travel-spend figure nor a net-worth-based HNWI measure. Conflating the two introduces a definitional error at the first step of the inferential chain.

2. How far apart are the figures that do have sources?

More consequential than an untraceable figure is the dispersion among the figures that are traceable.

2023, identical net-worth threshold (US$30m+), identical "ultra-high-net-worth" terminology:

A difference of 200,289 individuals, or 47%. The growth rates differ by a factor of nearly two.

2025, "millionaire" populations:

In a single year, the reported growth rate of the affluent population ranges from +1.5% to +14.4%, depending entirely on which publisher and which threshold one selects.

A substantial part of this dispersion is explicable, and is explained by the publishers themselves. Capgemini defines HNWIs as those holding investable assets of US$1m or more and explicitly excludes the primary residence. UBS defines net wealth to include owner-occupied housing and to deduct debt. These are simply not the same population.

Which is precisely the point. A statement of the form "global HNWIs grew X%" that does not specify threshold, asset definition and measurement year is, in a technical sense, undefined. It cannot be verified, cannot be compared, and cannot support any calculation requiring precision.

3. How these numbers are produced

To understand a 47% gap, one must look at how the figures are generated. The following descriptions are taken from each organisation's own published methodology.

Capgemini employs a proprietary two-stage model: national-accounts savings data from the IMF and World Bank are accumulated into a total wealth stock and adjusted to market value using global equity indices; the distribution across individuals is then derived from income-distribution data via a wealth-to-income relationship formula, across 71 markets.

Altrata / Wealth-X applies its Wealth and Investable Assets Model: econometric estimation from World Bank, IMF, OECD and national statistical sources produces total private wealth, which is then allocated across the population using Lorenz curves constructed from its proprietary database.

UBS describes its distributional estimate as "a model employing macroeconomic variables," covering 56 markets, including owner-occupied housing, deducting debt, and converting to US dollars at end-period exchange rates.

Henley & Partners / New World Wealth uses a country-benchmark model drawing on household income statistics, equity market capitalisation, an internal database, tax proxies and prime residential prices as a sanity check.

In other words, these high-net-worth population figures are not counts. They are model outputs — macro aggregates reallocated to individuals under a distributional assumption. That is not in itself a deficiency. In the absence of a global wealth census, it is the only tractable approach available.

What warrants attention are four disclosure conditions that follow, each drawn from the publishers' own documents:

None of the five organisations discloses a confidence interval or margin of error for its high-net-worth population estimate.

None of the estimates is independently audited or peer reviewed. (PwC Switzerland's role in the UBS report is described as data support, which is not an audit opinion on the results.)

Knight Frank states in its 2026 edition that "full methodology [is] available on request" — that is, not published in the report. Its data sources are listed as including Forbes. Its publication disclaimer states that the report "is not definitive and it is not to be relied upon in any way." The firm also notes that its model is dynamic and that figures "may not be identical to… previous editions," which means its population series should not be treated as a year-on-year comparable time series.

UBS states plainly that "it is primarily changes in foreign exchange rates that alter the relative performance of different economies' wealth," citing an approximately 9% appreciation of the euro against the dollar in 2025. The report does not, however, decompose how much of the change in the millionaire headcount is attributable to currency movement. For a population count defined by a US-dollar threshold, exchange-rate movement mechanically changes how many individuals sit above the line, so the absence of that decomposition is material.

It is worth noting that Henley & Partners / New World Wealth states explicitly in its methodology that its results should be read as "modeled estimates… illustrative indicators of broad trends… rather than as precise counts." That self-limitation is accurate, and it applies with equal force to every figure discussed in this section. The difficulty does not lie in what the publishers disclose. It lies in the fact that users of these figures discard the qualification when they place an "illustrative indicator of a broad trend" into a financial model as a precise parameter.

The underlying methodological problems are well documented in the academic literature. The under-coverage of the top tail in wealth surveys, and differential non-response among the wealthy, have been established repeatedly (Vermeulen, ECB, 2014; Bach, Thiemann and Zucco, DIW Discussion Paper 1717, 2018). Where Pareto or Lorenz-type distributions are used to extrapolate the top of the distribution, the choice of the tail-threshold parameter materially changes the resulting headcount (Eckerstorfer et al., Review of Income and Wealth, 2015). The accuracy of rich lists as a top-tail data source has its own dedicated literature (Wildauer and Heck, GPERC WP92; Raub, Johnson and Newcomb, IRS Statistics of Income, 2010).

These studies criticise the inherent difficulty of the measurement problem, not the integrity of any publisher. But together they establish one point: estimates of top-tail wealth populations carry non-trivial model uncertainty, and that uncertainty is currently not quantified in disclosure.

One qualification must be stated precisely. The bias direction demonstrated in this literature is understatement of top wealth, not overstatement. It therefore does not follow, from the public record, that commercial estimates run high. What can be asserted is that the uncertainty is undisclosed. It cannot be asserted that the figures are inflated.

4. Even if the figure is accurate: stocks and flows

Now grant the assumption. Suppose "15.5% growth in global HNWIs" is exactly right. Does luxury accommodation demand follow?

This is a question of elementary economics, and the answer turns on three conversion steps, none of which holds automatically.

First, a stock is not a flow. High-net-worth population is a stock — the number of individuals above a threshold at a point in time. Hotel demand is a flow — room-nights sold over a period. Deriving the second from the first requires a conversion coefficient: nights per person per year. That coefficient must be measured independently. It cannot be assumed.

Second, paper wealth is not purchasing power. These wealth aggregates include equity market values and property valuations. When asset prices rise, more individuals cross a dollar threshold, but that increment is an unrealised valuation change and does not necessarily convert into current discretionary expenditure. There is direct empirical evidence on this point. A study testing wealth effects on outbound tourism demand (Korea, 1989 Q1–2009 Q4, N=83, estimated with both Prais–Winsten FGLS and Newey–West OLS) found an income elasticity of 1.50 (t = 16.46, p < 0.01), a housing-wealth elasticity of 0.39 (t = 2.56, p < 0.05), and an equity-market wealth coefficient of −0.028 (t = −0.60), statistically insignificant. The authors report that the stock-market wealth-effect hypothesis was rejected in both models.

The significance of that result is this: the growth in high-net-worth headcounts reported by wealth studies is driven principally by equity market movements — Capgemini states explicitly that its total wealth is adjusted to market value using global equity indices. The variable driving the headline "more wealthy people" figure is precisely the variable that the available empirical evidence finds insignificant for travel demand.

Third, the units must be multiplicable. Suppose one accepts the calculation "HNWI headcount × nights per person per year = market size." Every available survey of nights per year among affluent travellers — for example the Luxury Institute's seven-country study, reporting roughly 19 hotel nights per year of which approximately 11 are in luxury properties — segments respondents by household income, not by net worth. Every HNWI population estimate segments by net worth. The two multiplicands describe non-overlapping populations, and the multiplication is therefore not defined.

There is also a more basic constraint. Hotel inventory is perishable, and individual consumption has a physical ceiling. One person occupies one room per night. Aggregate wealth has no such ceiling. There is consequently no identity linking the growth rate of wealth to the growth rate of room-night demand.

Finally, and stated as a finding rather than an argument: I found no published study testing the correlation between high-net-worth population and luxury hotel room-night demand. The causal chain is asserted narratively in industry research — one major advisory firm contrasts ten-year compound growth of 5.9% in millionaires, 5.2% in billionaires and 9.6% in global wealth against 2.3% compound growth in luxury supply, concluding that demand "exceeds supply growth" — but these passages contain no correlation coefficients, elasticity estimates or significance tests.

Two statements must be kept apart here. "This relationship has never been tested against published data" is accurate. "This relationship has been disproved" is not. I assert only the former. But for an asset class with a twenty-year-plus payback horizon, "never tested" is by itself sufficient grounds for caution.

5. Even if aggregate demand grows: why would it come to you?

Now suppose every concern above is set aside. The figure is accurate, the stock converts cleanly into a flow, and global luxury accommodation demand is genuinely growing. For any individual project, this still does not constitute an investment case.

This is the step most often skipped when macro narratives are applied to single assets: industry demand and firm-level demand are different curves.

If an ultra-high-net-worth individual takes five additional luxury hotel nights this year, those five nights will land at some subset of several thousand luxury properties worldwide. Where they land depends on destination choice, seasonality, air accessibility, brand relationship, existing loyalty, travelling-companion decisions, and above all on how many comparable rooms in that same destination are competing for those same five nights.

For a specific project, therefore, the correct question is not whether there are more wealthy people in the world. It is a sequence of progressively narrower questions:

Destination capture rate. What share of global ultra-luxury room-nights does this destination hold, and has that share risen or fallen over the past five years?

Local supply change. How many comparable rooms — comparable in ADR band, not merely in star rating — will enter this destination over the next 36 months? This is the denominator.

Project capture rate. Once that supply is fully absorbed into the market, what share can this property expect to capture, and on what basis?

Destination-specificity of demand. Does the destination offer a non-substitutable reason to visit? If not, why would the guest not choose a cheaper, more accessible alternative?

Seasonal concentration. Across how many weeks of the year is demand concentrated, and what covers fixed costs during the remainder?

Every one of these answers must come from measured data denominated in room-nights. None of them can be derived from any global wealth aggregate, however accurate. A macro figure, by construction, contains no information about the competitive structure of a destination.

The application of supply and demand here is unglamorous: rate and occupancy are determined by supply and demand within a specific market over a specific period, not by the global stock of wealth. Where comparable room supply in a destination grows persistently faster than room-nights actually sold in that destination, occupancy and achieved rate will come under pressure regardless of how many new millionaires exist globally. That is not a forecast. It is an identity.

One current, room-night-denominated observation illustrates the form the data should take. The Maldives is among the most established ultra-luxury accommodation destinations in the world. In 2025, bed capacity grew 4.3% while bednights sold grew 2.4%. By March 2026, with 4,818 additional beds year on year, resort bednights fell 12% and overall occupancy declined from 65% to 57%. The Maldives Monetary Authority series records resort occupancy at 50.1% in June 2026.

I draw no conclusion here about that destination's long-run prospects. Short-run movements in a single market have many possible causes and do not extrapolate globally. I cite it only because it is one of the few markets that publishes both sides of the supply-demand relationship in room-night units — and it therefore demonstrates the point that matters: in a market where supply growth persistently exceeds room-night growth, the direction of occupancy has nothing to do with global wealth statistics. That is the data structure an investor actually needs.

As for the very large sovereign-backed tourism developments in the Middle East now under discussion — including the reassessment of later phases of the Red Sea project — I would caution against treating them as a simple cautionary tale. Sovereign-wealth-backed projects pursue national strategic objectives, and their cost-benefit framework differs fundamentally from that of private capital. Their experience does not transfer to a project in which private or family capital bears the entire downside. If anything the inference runs the other way: precisely because a sovereign vehicle can absorb losses that no purely commercial structure could, its planning logic is of limited use as a reference for private investors. To replicate the scale without the risk-bearing structure is a category error.

6. The evidentiary standard feasibility studies should meet

The foregoing points toward a concrete and actionable set of requirements. They are directed at no particular organisation; they are general recommendations for luxury hotel feasibility work.

For every macro figure cited, require:

A primary source — publishing organisation, report title, edition year, page. A citation of a citation is not a source.

A precise definition — net worth or investable assets? What threshold? Is the primary residence included? Is debt deducted? In what currency, at what exchange rate, as of what date?

The measurement method — a count or a model estimate? If a model, under what distributional assumption?

The uncertainty — a confidence interval or margin of error. Where none is published, the document should state explicitly that the figure carries no uncertainty estimate.

A comparability statement — is the series comparable across years? Has the publisher restated historical values?

For every demand argument, require:

Demand denominated in room-nights or bednights sold — not wealth stocks, population stocks, loyalty membership counts, or unsegmented pipeline totals.

Supply and demand presented on the same basis — same destination, same ADR band, same time window, incoming supply against room-nights actually sold.

An explicit capture-rate assumption — the share the project expects to win, set out as a separately testable assumption rather than buried inside an aggregate derivation.

A distinction between industry demand and firm demand — growth in the former is not evidence of the latter.

A distinction between "untested" and "disproved" — where a causal chain has no supporting data, it should be labelled untested and the margin of safety adjusted accordingly, rather than left both unlabelled and untested.

Conclusion

This article asserts that no figure has been fabricated and alleges no misconduct by any organisation. On the contrary: in the absence of a global wealth census, estimating high-net-worth populations from macro aggregates and distributional assumptions is a genuinely difficult technical undertaking, and the methodological effort involved deserves respect. One publisher states in its own methodology that its results are "illustrative indicators of broad trends… rather than as precise counts." That is a commendable piece of candour.

The failure occurs in transmission. A model output, honestly labelled as an illustrative trend indicator, passes through several rounds of secondary citation. Along the way it loses its definition, its measurement basis and its uncertainty qualification. It arrives on the first page of an investment memorandum as an isolated percentage and is then treated as a precise parameter in a twenty-year underwriting model.

No one lies at any point in that process. The conclusion is nonetheless unreliable.

The recommendation I would offer is therefore a measured one: in luxury hotel development decisions, return macro wealth data to its proper role. It is context, not evidence. What can support an investment decision is destination-level, room-night-denominated data available on both the supply and demand sides; an explicit capture-rate assumption; and a specific answer to the question of why these guests will come to this destination and stay at this property.

Rigour about data is not academic fastidiousness. It is the margin of safety on the capital. Where a figure cannot produce a primary source, no calculation built upon it can be more precise than the figure itself.

Sources

Altrata, World Ultra Wealth Report 2026: https://altrata.com/wp-content/uploads/2026/06/Altrata_World-Ultra-Wealth-Report-2026_FINAL.pdf

Altrata, World Ultra Wealth Report 2024: https://altrata.com/reports/world-ultra-wealth-report-2024

Capgemini, World Wealth Report 2026: https://www.capgemini.com/de-de/wp-content/uploads/sites/8/2026/06/2026-05-26-STUDIE-World-Wealth-Report-2026-1.pdf

Capgemini, World Wealth Report research library: https://www.capgemini.com/insights/research-library/world-wealth-report/

UBS, Global Wealth Report 2026: https://www.ubs.com/content/dam/assets/wm/static/gwr/global-wealth-report-en-2026.pdf

Knight Frank, The Wealth Report 2024: https://www.knightfrank.com/site-assets/research/reports/the-wealth-report/previous-editions/the-wealth-report-2024.pdf

Knight Frank, The Wealth Report 2025: https://apac.knightfrank.com/hubfs/Research Reports/Residential/Report PDFs/Knight Frank_The Wealth Report 2025.pdf

Knight Frank, The Wealth Report 2026: https://www.knightfrank.fr/fichiers/publications2020/file//146815-the-wealth-report-2026-6a2298f435832632958583.pdf

Henley & Partners / New World Wealth, Methodology: https://www.henleyglobal.com/publications/africa-wealth-report-2025/methodology

Skift, Media Resources: https://skift.com/media-resources/

Skift Research, A Deep Dive Into Luxury Hotels: https://research.skift.com/reports/a-deep-dive-into-luxury-hotels/

Marketplace, "Higher-income Americans drive bigger share of consumer spending" (Moody's Analytics for WSJ): https://www.marketplace.org/story/2025/02/24/higher-income-americans-drive-bigger-share-of-consumer-spending

Federal Reserve Bank of Minneapolis, "Have US consumers gone K-shaped? A review of the data": https://www.minneapolisfed.org/article/2026/have-us-consumers-gone-k-shaped-a-review-of-the-data

JLL, 2026 Global Hotel Investment Outlook: https://www.jll.com/content/dam/jllcom/en/global/documents/reports/research-reports/26-insights-global-hotel-investment.pdf

Vermeulen, P. (2014), "How fat is the top tail of the wealth distribution?", European Central Bank: http://www.piketty.pse.ens.fr/files/Vermeulen2014.pdf

Bach, S., Thiemann, A. and Zucco, A. (2018), "Looking for the Missing Rich", DIW Discussion Paper 1717: https://www.diw.de/documents/publikationen/73/diw_01.c.575768.de/dp1717.pdf

Eckerstorfer, P. et al. (2015), "Correcting for the Missing Rich", Review of Income and Wealth: https://jakob-kapeller.org/images/pubs/2015-Eckerstorferetal-ROIW.pdf

Davies, J., Sandström, S., Shorrocks, A. and Wolff, E. (2011), "The Level and Distribution of Global Household Wealth", The Economic Journal: http://piketty.pse.ens.fr/files/DaviesEtal11.pdf

Wildauer, R. and Heck, I., "Was Pareto right?", GPERC Working Paper 92: https://gala.gre.ac.uk/id/eprint/38597/13/38597 WILDAUER_Was_Pareto_right_Is_the_distribution_of wealth_thick_tailed_(REVISED)_2023.pdf

Raub, B., Johnson, B. and Newcomb, J. (2010), IRS Statistics of Income: http://piketty.pse.ens.fr/files/RaubJohnsonNewcomb2010.pdf

Alvaredo, F., Berman, Y. and Morelli, S. (2024), "Evidence from the Dead", IZA Discussion Paper 17389: https://docs.iza.org/dp17389.pdf

"Wealth Effect and Demand for Outbound Tourism", UMass ScholarWorks: https://scholarworks.umass.edu/bitstreams/74d6d59b-61fd-4b06-af1c-c105ef105b7b/download

Luxury Institute, Global Hotels LBSI survey release: https://www.einpresswire.com/article/299666196/luxury-institute-survey-provides-country-by-country-rankings-of-global-hotel-brands-by-affluent-travelers-from-the-world-s-richest-countries

Visit Maldives, Quarterly Insights Q1 2026: https://corporate.visitmaldives.com/news/quarterly-insights-q1-2026-tourism-performance-and-demand-outlook/

Corporate Maldives, local island tourism and arrivals coverage: https://corporatemaldives.com/local-island-tourism-holds-ground-amid-tourist-arrival-slump/

Maldives Monetary Authority, resort occupancy series: https://database.mma.gov.mv/viya/series/219