On the Evidentiary Basis for Luxury Hotel Feasibility

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:

1. What is the primary source of this figure?

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

3. 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:

Source	Population	Growth that year
Knight Frank (Wealth Sizing Model)	626,619	+4.2%
Altrata / Wealth-X (Wealth and Investable Assets Model)	426,330	+7.6%

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

2025, “millionaire” populations:

Source	Population	Growth that year
UBS Global Wealth Report 2026	58,000,000	+1.5%
Capgemini World Wealth Report 2026	25,300,000	+7.9%
Altrata (UHNW, US$30m+)	556,850	+14.4%

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:

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

2. 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.)

3. 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.

4. 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:

1. 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?

2. 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.

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

4. 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?

5. 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:

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

2. 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?

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

4. 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.

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

For every demand argument, require:

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

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

8. 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.

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

10.  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

About the author

Tong Yin, Ph.D., holds a doctorate in hospitality management from Auburn University and is the founder of InsightBridge Global LLC. His research and consulting work focus on ultra-luxury hotel asset management, organizational behavior, and the evolving business model of international hotel groups.

[email protected] · insightbridge.global