RESEARCH DOSSIER

Measurement of Offshore Household Wealth

Abstract. This research dossier investigates the measurement of offshore household financial wealth through international investment statistics, bank balance sheets, ownership allocation, tax administrative data, and asset-specific studies. It assesses global securities discrepancies, deposit-based proxies, custodial and entity look-through, distributional adjustment, automatic information exchange, causal inference, stock-flow reconciliation, valuation, and uncertainty. Published estimates are evaluated in their original periods, asset perimeters, and denominators. The original analytical synthesis distinguishes economic ownership, statistical recognition, and fiscal disclosure, and uses explicitly hypothetical arithmetic to demonstrate the consequences of their conflation. No new empirical stock estimate, named theory, or independent econometric replication is claimed.

Keywords: offshore household wealth; offshore financial assets; international investment positions; household balance sheets; financial intermediation; ultimate ownership; portfolio discrepancies; locational banking statistics; tax transparency; wealth inequality; measurement uncertainty

Introduction

A foreign financial position has several geographic meanings. It can belong to a household resident in one economy, be administered by a bank in a second, be issued by an investment fund incorporated in a third, and finance an enterprise located in a fourth. A single geography is adequate only for a precisely specified accounting question. The measurement of household offshore wealth begins by deciding which of these relationships constitutes the object of inquiry. Otherwise a statistic about where securities are held becomes a statistic about who owns them, and a statistic about who owns them becomes an assertion about tax compliance.

This research dossier examines the measurement of stocks belonging economically to households and held through offshore financial arrangements. Its chief question is how evidence collected for different institutional purposes can be combined into a consistent estimate of ownership without confusing intermediaries with principals. The investigation connects external-sector accounts, banking data, confidential administrative records, balance-sheet statistics and asset-specific evidence. It reviews published quantitative results and develops accounting examples and methodological assessments independently of those results. It does not supply a newly estimated country panel. The examples use invented amounts and named variables solely to expose the mechanics of measurement.

The empirical literature grew out of a striking discrepancy in global financial accounts. Zucman's (2013) The Missing Wealth of Nations connected this discrepancy to household portfolios held through foreign institutions. Subsequent research extended the allocation of these assets across owners, investigated their distribution among households, and examined disclosure reforms. These are related but distinguishable empirical tasks. The total stock can be estimated even when its precise country allocation is uncertain, while the tax status of an account can be measured much more accurately in a single administrative jurisdiction than globally.

The scientific contribution of the dossier is a critical synthesis of identification problems across the measurement sequence. The organizing conclusion is that ownership reconstruction, statistical coverage, and declaration status must be maintained as separate dimensions until they can be jointly observed. The meaning of each estimate follows from its denominator, asset perimeter, reference date, and attribution rule. This is an analytical conclusion from the comparison of established methods, not a new named theory.

The Object of Measurement

A household offshore financial asset, for the purposes of this dossier, is an economically attributable claim held through a foreign financial intermediary or comparable offshore arrangement under a specified geographic perimeter. The perimeter must be given explicitly. A broad definition encompasses any foreign account or investment arrangement used by resident households; a narrower empirical definition selects a set of financial centers and particular account or portfolio arrangements. Neither is universally interchangeable with the other. A remote brokerage account in an ordinary foreign economy can satisfy a broad cross-border definition without entering a narrower offshore-center series.

Ownership requires a unit of account. Individual tax records can be aggregated into households only under an explicit grouping rule. Household sector accounts may encompass individuals conducting unincorporated business, and some banking series combine households with nonprofit institutions serving households. The latter convention is explicit in the BIS locational banking glossary. Household ownership through an incorporated enterprise requires a separate decision: the enterprise's assets are its own balance-sheet positions, whereas household shareholders own equity claims against it. Any transparent look-through must eliminate the intermediate claim before replacing it with underlying assets and liabilities.

A gross financial stock measures the fair or relevant reporting value of included claims. Net financial wealth deducts financial liabilities within the adopted boundary. A household with a foreign securities portfolio worth 1 million dollars and a directly associated debt of 300,000 dollars has 1 million dollars of gross assets and 700,000 dollars of net exposure under this simple example. Another household may have offshore assets financed by domestic debt; whether the debt is assigned to the offshore subtotal is a methodological choice. Inconsistent liability treatment makes a comparison of 'wealth' between countries misleading.

There are at least three distinct observational questions. Was the economic claim captured by the national statistical compilation? Was the asset or relevant income disclosed to the responsible tax authority where a duty existed? Where was the service, custody, or legal vehicle located? A properly declared investment may be absent from a statistical survey, while an asset that is included in aggregate economic statistics may remain unidentified for fiscal purposes. The aggregate statistical system generally need not identify every household individually. Accordingly, neither statistical invisibility nor offshore location establishes a legal violation.

The target population should be determined by economic residence at the reference date, rather than nationality, passport, ethnicity, or place of birth. Tax residence should be recorded separately where available. An individual who changes residence may move into or out of a country's measured stock with no transfer of money. Joint tax residence, disputes about residence, discretionary trusts and ownership structures involving multiple households complicate operational assignment. A measure of nationals' wealth abroad is a different estimand and may be useful for another purpose, provided the label makes it explicit.

International Accounts and the Global Portfolio Discrepancy

In an internally consistent global financial ledger, one unit's cross-border portfolio liability corresponds to another unit's cross-border asset. At the world level, properly matched claims and liabilities at the same valuation date should cancel. Actual reporting populations, classifications, methods and timing differ. Summing reported positions therefore creates a residual. Let L denote the aligned sum of recorded portfolio liabilities and A the corresponding sum of recorded portfolio assets. The observed discrepancy G = L − A is a statistical signal, not an observation of a particular hidden account.

The IMF Portfolio Investment Positions dataset collects international portfolio holdings on a voluntary basis, with distinct instrument groups for equity and investment fund shares, long-term debt and short-term debt. Its value for an offshore wealth study is the possibility of comparing asset-side records and supplementing them with other external statistics. The required population alignment is not automatic. Coverage of reporting economies, international organizations, valuation procedures, asset categories, survey methods and confidentiality suppressions must be reconciled before treating the difference as informative.

Suppose a resident of economy A owns 100 dollars of securities issued in B, held in custody in C. B reports a 100-dollar external liability. If A misses the household's holding, C's custodian does not acquire a beneficial asset merely by safeguarding it. The global accounts show a 100-dollar liability with no corresponding asset. In a perfectly matched ledger, A records the position against B even when C administers it. The identity illustrates how an ownership visibility problem can generate an asymmetry without making custody an investment by the custodian itself.

The same residual could appear for an entirely different reason. If a sovereign investor fails to report a portfolio, the ownership gap has nothing to do with resident households. If asset data are collected in book values while liabilities reflect market value, both sides can be recorded and still disagree. A change in the reporting treatment of investment funds may produce a discontinuity. Asset-side undercoverage and liability-side overstatement both increase G; liability undercoverage and asset overstatement push it in the opposite direction. Offsetting errors can conceal a considerable true coverage problem even when G is small. The direction of bias from the raw residual is therefore indeterminate.

Write the analytical decomposition G = H + O + E, where H is the missing household portfolio component of interest, O is other net undercoverage, and E is residual valuation, classification and measurement error. This equality defines the components conceptually; it supplies no independently identified values. Estimation requires auxiliary evidence about O and E. If G is 120 and known net nonhousehold missing assets plausibly range from 15 to 45, with remaining net error between −10 and +10, H lies between 65 and 115 under those assumptions. The intervals are scenario bounds, not probability-based confidence intervals. Negative component estimates require revisiting assumptions rather than mechanically taking absolute values.

Zucman's (2013) inference combines the global discrepancy with Swiss evidence to give the residual economic interpretation. The strength of the approach is aggregate identification through accounting consistency and an independent institutional source. Its limitation is that the inference depends on the completeness of correction terms and the correspondence between missing portfolios and the target population. It estimates a latent quantity. It cannot identify individual owners or infer their motives from the balance of international accounts.

Deposits and Custodial Securities

Bank deposits and securities held by a bank for a client have different accounting treatments. A deposit is a financial claim against the bank. A security held in custody ordinarily remains a claim against the issuing unit; the intermediary supplies a service. International banking statistics provide a complementary lens because deposits are on the intermediary's balance sheet, whereas ordinary custody assets are not bank liabilities merely by being administered there.

The BIS Locational Banking Statistics organize international banking positions around the reporting banking office and counterparty residence; their glossary also explains sector groupings and adjustments for exchange rates and reporting breaks. The liability of a bank to a nonresident provides information about a corresponding claim held abroad. A foreign office of a bank headquartered domestically is a foreign reporting office for locational purposes; the identity of the parent banking group answers a different question. Mixing national bank ownership with the location of its balance sheet obscures the actual geographic boundary.

Deposits must not be added indiscriminately to portfolio securities. Consider 100 dollars in a household's brokerage cash balance, temporarily deposited by the broker with a bank. A ledger may contain the customer's claim against the broker and the broker's deposit claim against the bank. Counting both as distinct household assets would inflate the estimate. If the balance is instead a genuine deposit held directly in the household's name and the household also owns 200 dollars of independently acquired bonds, gross financial assets are 300 dollars. The difference resides in the chain of economic claims.

Some studies use a benchmark for the relative importance of deposits and securities to infer a total offshore financial portfolio. If estimated missing securities are 70 and one assumes securities comprise 70% of the target total, the implied 100 depends on the meaning of the 70% factor. Where the benchmark describes all offshore securities, including recorded positions, estimating total wealth from only missing securities requires an additional coverage parameter. Suppose 80% of all offshore securities are missing from the external statistics; 70 units of missing securities then imply total securities of 87.5, and if these constitute 70% of holdings, total wealth is 125. This illustration separates the coverage correction from asset composition.

For another example, suppose center X primarily manages securities and center Y mainly receives deposits. Equal deposits of 10 in each do not imply equal total wealth under a conversion based on a common securities ratio. If deposits make up 20% at X and 50% at Y, corresponding totals are 50 and 20. This is why a portfolio composition ratio should be estimated for specific center-period cells where feasible, varied systematically where unavailable, and recorded with a source and justification.

Researchers using bilateral positions also face sector detail limits. A nonbank category contains more than households, and incorporated intermediaries can be recorded in other sectors even when households ultimately own them. Restricted supervisory breakdowns can sharpen sector identification, but the resulting estimate cannot be validated from public aggregates alone. A discrepancy between an aggregate banking series and household administrative reports can reflect sector composition or data coverage, rather than a change in household wealth. An asset estimate should name the exact instrument and source population instead of using 'offshore deposits' as a general synonym.

From Financial Centers to Ultimate Household Owners

Let H_j be a reconciled household offshore financial stock administered in center j, and p_ij be the fraction economically attributable to residents of owner country i. If all owners can be assigned, the column sum of p_ij equals one and the estimated country total is W_i = sum over j of p_ij H_j. In practice the allocation may contain an unassigned category. Leaving that category visible is preferable to forcing weakly identified cells to absorb the remainder.

Alstadsæter, Johannesen, and Zucman's (2018) country distribution of offshore wealth connects estimates of the global total with country-specific ownership proxies. The result enables comparison of origins, rather than merely the jurisdictions hosting financial institutions. A country-level distribution inferred from banking data does not imply that the owner of every individual security has been observed. It transfers an observed ownership pattern to an unobserved population, and the accuracy of the transfer depends on similarities in the types of clients and instruments represented.

Consider two centers with estimated stocks of 60 and 40. Suppose proxy allocations assign 75% of the first to country A and 25% to country B; the second is split 25% and 75%. A receives 55 units and B 45. If A's first-center share lies instead between 60% and 80%, with the second share fixed, the result for A spans 46 to 58. The error varies by the center's importance to A, even though the common global total remains 100. Equal uncertainty about each share has unequal consequences for different countries.

Intermediary residence can itself be another center. The bank's immediate claimant may be an incorporated vehicle whose shareholders or beneficiaries live elsewhere. Suppose a center records 20 units against entities resident in an intermediate jurisdiction. The evidence may identify half as economically owned by A, a quarter by B and leave one quarter unknown. Reassigning all 20 proportionally to the identified subset would convert missingness into apparent precision. A consistent allocation would move 10 to A, 5 to B and leave 5 unassigned unless additional evidence supports distribution of the remainder.

A fund registration is particularly important because its domicile may determine the debtor economy for fund shares while its portfolio contains claims against several other economies. The estimate of household claims against the fund should not be added to a separate estimate of the same household's underlying look-through exposure. The first preserves legal instrument geography; the second preserves ultimate investment exposure. Beck and colleagues' (2024) Euro area capital allocation research demonstrates the importance of both fund-holder and issuer look-through for interpreting portfolio geography. The methodological lesson is to maintain separate accounting perspectives and a documented bridge between them.

A residence reassignment matrix is not a moral classification of countries. A large hosting center may administer assets belonging to many nonresidents and have comparatively little resident-owned offshore wealth; an owner country may have very little domestic offshore-service activity and a large externally administered stock. These two roles should be reported in separate tables. A third table may describe the location of underlying investments if the information is available. Using a single 'offshore country' ranking collapses fundamentally different positions in the network.

Evolution of the Global Estimates

The foundational empirical estimates need to be read with their original denominators and vintages. Zucman (2013) estimated that about 8% of global household financial wealth was held in havens, approximately three quarters of it absent from official external statistics. Alstadsæter, Johannesen, and Zucman (2018) found offshore wealth on the order of 10% of world GDP and substantial differences between resident countries. The denominator in the first statement is a wealth stock; in the second it is annual economic output. They do not represent two observations of the same ratio.

The 2001–2023 offshore wealth series of Faye, Godar, Moura, and Zucman (2025) puts 2023 offshore financial wealth near USD 13 trillion and describes relative stability over the period together with marked changes in hosting and origin geography. It is a working paper and an updated methodology, rather than a final peer-reviewed census. Its historical series corrects coverage problems using auxiliary data and revises the distribution across centers. The long-run result is therefore evidence within a consistent reconstruction, not a sequence of directly observed household portfolios.

The European Commission and ECORYS (2021) offshore wealth monitoring report reports a financial wealth estimate of USD 9.8 trillion for 2018 and considers other asset classes separately. A comparison with a later working paper's stock can usefully indicate scale, but different estimation vintages and boundaries prevent reading the difference as a directly measured transfer.

Three distinctions govern interpretation of these findings. First, estimates that employ common portfolio and banking inputs share errors. Agreement is a robustness signal only to the extent that independent assumptions and evidence contribute to it. Second, a nominal stock in US dollars is sensitive to market valuation and the exchange rates applied at the reference date. Third, stability relative to GDP is compatible with rising nominal wealth and with changing owner populations and levels of disclosure. The combination of aggregate continuity and geographic change is particularly informative because it implies that compositional movements cannot be read from a single global time series.

A publication should give the observation year as well as the publication year. A paper published in 2025 whose last observation is 2023 does not establish the 2025 or 2026 level. The evidence reviewed here supports the magnitude and interpretation of the published observations; it does not justify extrapolating them as current data without a specified extension method. Reporting an unqualified live global total based on historical estimates would disguise the principal source limitation.

Administrative Records, Leaks, and Wealth Distribution

Macroeconomic stock reconstruction has limited information about the distribution among persons. Tax returns, administrative balance-sheet files, financial reports and disclosures can help estimate concentration. Their advantages are greatest where records can be securely matched to an independently measured resident population. A file containing many customers of an offshore institution can reveal patterns in that institution's client base while still providing no representative sample of all residents.

Alstadsæter, Johannesen, and Zucman's (2019) Tax Evasion and Inequality compares evidence from offshore leaks, tax amnesties, random audits, and Scandinavian population records. The combination is important because detection methods have distinct selection effects. Random audits can be representative at the selection stage while failing to detect sophisticated holdings. Leaks may reveal otherwise invisible positions but represent particular providers, dates and clientele. Amnesty participation depends on incentives and perceived enforcement. No individual source is a universal sample of undisclosed wealth.

Let H_g be the aggregate offshore wealth assigned to wealth group g. The population-level offshore stock is the sum of H_g, and the group share is H_g divided by that total. The probability of owning any offshore assets is a different statistic from the average asset value conditional on ownership. A small group can account for large aggregate holdings through either a high participation rate, high conditional values, or both. Estimates of concentration should report these components where available.

Suppose a national wealth distribution has total measured net wealth of 100 and a richest group holding 25. If 10 units of previously omitted wealth are attributed exclusively to that group, the corrected share is 35/110, or approximately 31.8%. If the same 10 is dispersed in proportion to previously observed wealth, the top share remains 25%. Both assumptions produce the same corrected aggregate. The inequality conclusion comes from distributional allocation, not from the existence of an aggregate omission.

The starting measure must also be specified. A foreign account may already be reflected in survey assets or national accounts even when it is absent from a tax return. Correcting a wealth distribution using offshore data requires the incremental amount missing from the wealth dataset, rather than the entire balance reported by a foreign institution. If public aggregate wealth is already benchmarked using an external statistical estimate, adding the complete reconstructed offshore stock may double count it.

Colombia provides a useful contrast to high-income administrative settings. Londoño-Vélez and Ávila-Mahecha's (2021) Colombian wealth tax study uses a disclosure initiative and related evidence to document concentrated hidden wealth and policy responses. Different enforcement institutions, levels of wealth, financial access and the selection into disclosure matter for external validity. A common qualitative conclusion about top-tail concentration does not establish an identical distributional correction for every economy.

The measurement of wealth rankings requires special care because adding missing assets can change a person's group assignment. Applying offshore estimates to deciles defined before the correction and reporting shares as if group membership were based on corrected wealth introduces an inconsistency. Where microdata permit, researchers should recompute ranks under the final wealth definition and report the effect of re-ranking separately. Where only group-level estimates exist, distributional changes depend on assumptions about rank transitions. The problem is most severe around thresholds used to define very small top groups.

Automatic Exchange and Tax Visibility

The policy significance of offshore wealth depends in part on what authorities know. The OECD (2025) consolidated Common Reporting Standard describes an annual exchange framework under which participating jurisdictions obtain information from reporting financial institutions and exchange it with partner authorities. Amendments broaden specified product coverage and strengthen reporting and due diligence. Actual reporting depends on the applicable implementation, exchange relationships, effective dates and exceptions; a published model text does not establish identical practice everywhere.

The OECD reporting and due diligence rules include defined terms for relevant persons and entities and account aggregation procedures. In particular, rules may attribute full joint balances to each holder for reporting-related aggregation. Such records should not simply be summed across recipients as if every person owned a separate full economic claim. A household estimate requires deduplication and, where the inquiry calls for individual attribution, an appropriate allocation of joint economic interests.

A data exchange can make a previously existing asset visible without creating any new asset or flow. An account can enter the receiving authority's database because its owner first declares it, because information is transmitted independently, or because the records are matched after a delay. The number of reported accounts, the value of reported accounts, and the incremental value recognized by the tax authority measure different outcomes. An increase in reported wealth may signal an improvement in measurement rather than increasing offshore activity.

Boas and colleagues (2024), in Taxing Capital in a Globalized World, study Danish matched foreign-account reports, administrative records and bank transfers. They distinguish return of assets, voluntary reporting and detection through audit. This research design clarifies why deposit movements alone are insufficient for measuring policy performance: increased disclosure can occur without withdrawal or closure. The institutional context, accessible population and data linkage capacity constrain generalization.

A useful statistical representation distinguishes four cells: recorded and declared; recorded and undeclared; unrecorded and declared; unrecorded and undeclared. Only a linked dataset can directly assign claims along both dimensions, and the legal applicability of declaration duties must first be determined. Knowing that 70 of 100 units are statistically recorded and 80 of 100 are fiscally declared does not establish that 70 units are both. Under those margins the joint cell can range from 50 to 70. The unrecorded and undeclared intersection correspondingly ranges from zero to 20. Separate aggregate margins cannot establish the amount hidden from both systems.

Even a well-functioning automatic exchange system provides an administrative view of covered accounts, not a definitive census of worldwide asset ownership. Financial product coverage, institutional exclusions, participation, identification quality, privacy rules and effective dates create boundaries. These issues call for validation and transparent aggregation rather than access to personally identifiable account information. Researchers should use pseudonymized linkage environments, independent authorization and disclosure controls. The policy goal of transparency coexists with legitimate confidentiality and data protection.

Identifying the Effects of Transparency Reform

A change in offshore holdings following a reform can reflect asset transfers, declaration, valuation, reclassification, participant selection or contemporaneous economic shocks. A causal effect compares the change to a credible counterfactual. Before-and-after comparisons are particularly vulnerable when treaties come into force around global events or changes in asset prices.

Bénétrix, Emter, and Schmitz (2024) study automatic information exchange using restricted BIS sectoral information and bilateral investment data. Their estimates for the 2014–2019 setting distinguish the larger decline in relevant household deposits from that of the broader nonbank aggregate. The design's value lies in separating depositor sectors and comparing treaty timing; the published result of roughly 28.5% for households versus 12.5% for nonbanks refers to the specified bilateral deposit response, not a global loss of household assets. It also documents evidence consistent with geographic substitution.

Consider an event study in which Y_ijt denotes the log of deposits of households from i in center j at time t. A coefficient for time relative to treaty signature may capture changes around the event after specified fixed effects and controls. Identification requires defensible comparison pairs, anticipation treatment, common-shock adjustments, robust inference at the appropriate clustering level and a clearly defined treatment date. Treaty announcement, signature, effective reporting period and first exchange are often economically different events. The coefficient also depends on how zero balances, censored values and revisions are treated.

Staggered treatment complicates comparisons because early treated units may cease to be valid controls for later treated units when effects vary across time or groups. Researchers should test the identifying assumptions using pre-event estimates, alternative comparison sets and sensitivity to placebo dates. These are diagnostic requirements for a statistical design, not evidence that a specific published study fails them. Effects on uncovered centers can violate a simple no-interference assumption. The geographic displacement of accounts may itself be an economically meaningful outcome.

A successful tax compliance policy should be judged against the appropriate outcome. The stock of foreign assets may remain stable while the amount of undeclared wealth declines. Deposit outflows from participating centers may be accompanied by increased holdings elsewhere. Annual tax revenues can rise because authorities detect income even when portfolio values remain unchanged. A macro stock series, a bilateral deposit event study and a matched compliance study provide complementary information on these processes.

The Global Tax Evasion Report 2024 of Alstadsæter and colleagues (2023) situates offshore household assets within a broader set of information-based reforms. Its scope includes activities with different legal characteristics. As an analytical matter, the difference between legal foreign asset holding and unlawful non-disclosure is necessary for evaluating a policy's intended effects. The measurement problem belongs to economics; the determination of an individual's legal obligations and violation belongs to jurisdiction-specific law and fact.

Stocks, Transactions, and Asset Valuation

A stock at the end of the year measures the value at a reference instant. A transaction measures a transfer or exchange over an interval. Let W_t denote offshore wealth at date t, F_t net financial acquisition through transactions, Q_t price revaluation, X_t currency revaluation and O_t other volume changes, including statistically recognized changes in coverage or residence. Then W_t − W_(t−1) = F_t + Q_t + X_t + O_t, subject to the exact accounting conventions used. Transfers between centers may cancel at the global level yet change the host-country distribution.

Suppose an illustrative stock rises from 100 to 125. Of the increase, 14 reflects higher asset prices, 5 reflects exchange rates, and 4 reflects the inclusion of previously unmeasured holdings. The implied net acquisition is only 2. This calculation does not assert that the 4 newly measured units represent real capital movement. If the coverage change is backcast instead of recognized in the current period, the historical stock series itself may be revised.

The converse also holds. Suppose households withdraw 15 in net transactions while remaining securities appreciate by 20. The stock increases by 5 even though there was a net withdrawal. A simple before-and-after comparison would misclassify the direction of transactions. The BIS publishes exchange rate and break adjustments for certain banking positions; those adjusted changes are designed to approximate flows after isolating specific nontransaction changes, subject to their series definitions. Offshore portfolio wealth requires its own price, currency and membership reconciliation.

The valuation conventions for assets differ. A listed security usually has a market quotation at the valuation date. An unlisted entity may need a model-based valuation, and the balance of an insurance or pension claim may follow contract or actuarial rules. Value attributable to households through a vehicle should account for the vehicle's liabilities under the chosen equity or look-through treatment. Nominal account balances and marked-to-market financial positions can coexist in an aggregate only if their measurement bases are appropriate to the respective instruments and consistently dated.

Ratios introduce an additional temporal and unit distinction. W/GDP compares a year-end asset stock with annual production and has the interpretation of a scale relative to a year's output. W/household financial wealth compares two stocks. One country can have a higher offshore-to-GDP ratio and a lower offshore-to-financial-wealth ratio than another because its financial wealth relative to output is larger. For example, stocks of 20 for financial wealth of 200 and GDP of 50 yield ratios of 10% and 40% respectively; stocks of 15 for financial wealth of 100 and GDP of 100 yield 15% and 15%. The ranking reverses.

An estimate of concealed assets also cannot be converted to tax revenue by treating the asset principal as ordinary taxable annual income. A stylized income-tax calculation requires an assumed return, legally applicable rate, taxable share, relevant deductions and credit treatment. Suppose a previously undeclared asset of 100 produces taxable net income of 4, an illustrative effective tax rate of 25% would yield 1 under those assumptions. The estimate excludes penalties, taxes on prior years, wealth taxes or repatriation consequences. It is an illustration of the separation between stock, income and assessment, not a calculation of any actual country's revenue.

Alternative Asset Classes and Measurement Limits

The financial perimeter is empirically convenient because it can be investigated with banking, investment-position and administrative data. It does not encompass every form of household wealth held outside the residence economy. Physical real estate, art, cash outside institutions and other tangible assets have different ownership and valuation records. Some life insurance and investment products represent financial claims and may overlap with existing totals. Expansion of the perimeter requires a deduplication rule before any aggregate is presented.

Alstadsæter and colleagues' (2025) Dubai offshore real estate study provides a separate local property estimate using ownership microdata. Their revised study assigns nonresident owners an estimated 30% of residential property value in 2019, approximately USD 68 billion. This is asset- and jurisdiction-specific, and the survey of the source here is confined to the institutional abstract. It should not be multiplied by the population of other property markets or added to an earlier financial series without aligning entities, dates and valuation.

The ECORYS report's treatment of additional assets highlights another issue: financial and nonfinancial assets held through the same vehicle can appear both as underlying property and as equity wealth. A property valued at 100 and financed by debt of 40 creates 60 of net equity under the simplified assumption that those are the only assets and liabilities. A household owning the entire property company owns equity worth 60. Adding the property value of 100 and company equity of 60 as though both were unconnected household wealth counts the same economic interest twice. A coherent gross look-through presents 100 and separately 40 of associated liabilities; a claim-level approach presents 60 of equity.

Certain ownership categories are also conceptually different. A trust beneficiary with a fixed, enforceable entitlement has a different measured claim from a discretionary beneficiary whose entitlement is contingent. A person exercising control may not possess the full economic benefit. Administrative beneficial ownership definitions serve specified disclosure purposes and may include persons whose legal status does not imply ownership of every unit of value. An econometric allocation should document which persons are treated as economic beneficiaries and represent unresolved shares explicitly.

Physical currency held directly creates a further geographic question: the issuer of a banknote, its physical location and its owner's residence may all differ. Cash in a foreign safe deposit box is not generally recorded as a customer deposit merely because the box is located in a bank. A valuation method for these assets needs evidence other than deposit liabilities.

A broad offshore asset inventory is valuable, but components should be separately published when the linkage required to consolidate them is absent. A narrow and precisely measured financial stock can be more useful for international investment statistics than an apparently comprehensive total built from incompatible sources. Claims about complete worldwide wealth coverage are unwarranted until instrument, entity and ownership overlaps can be evaluated.

Uncertainty, Sensitivity, and Validation

Offshore wealth estimates involve several forms of uncertainty. Sampling variability can arise in household surveys or administrative samples. Measurement error enters through misclassification, valuation and reporting omissions. Model uncertainty enters through the choice of allocation proxies, composition ratios and residual corrections. Coverage uncertainty enters where centers or instruments have no observation. These have different statistical meanings and should not be collapsed into a mechanically calculated confidence interval.

A reproducible sensitivity exercise first defines the benchmark, then varies assumptions whose uncertainty affects the estimate. Suppose a corrected missing securities position is 80, the estimated fraction of all offshore securities absent from the relevant portfolio statistics ranges from 0.7 to 0.9, and the securities share of offshore financial portfolios ranges from 0.6 to 0.8. Under the strictly conditional expression W = 80/(missing fraction × securities share), the lower endpoint is 80/(0.9 × 0.8) = 111.11, while the upper endpoint is 80/(0.7 × 0.6) = 190.48. The endpoints assume the extreme parameters can occur jointly. If they cannot, the interval must reflect their dependence. The example uses no actual data.

This type of exercise also reveals how sensitive the estimate is to incomplete statistical capture: changes in the denominator are nonlinear when the coverage fraction is small. A center whose official data are sparse contributes not only a possibly large component but also a large uncertain amplification of that component. More precise data on small, well-measured centers cannot solve that identification problem. The research priority should reflect the contribution of a cell to uncertainty in the final estimate.

Validation requires independent information where possible. A global portfolio discrepancy can be compared with a separate host-administrative estimate, and country allocations can be compared with tax authority reports or representative wealth records where access is lawful. Agreement between results that use the same securities liability data and the same deposit proxies is weaker evidence than convergence of independently collected records. A ledger of source lineage should identify common inputs explicitly.

Longitudinal validation requires that revisions not be confused with changes in economic behavior. For an unchanged reference year, a shift between old and new estimates reveals differences in data vintage, corrections or assumptions. A revised historical time series should preserve the earlier vintage, explain substantive revisions and identify whether they concern observations or imputations. Missing values need explicit flags; zero should denote an economically meaningful zero only where the reporting convention permits that interpretation.

For a country allocation, one should inspect the sum of assigned and unassigned holdings against the independent global constraint. If estimates for individual centers are themselves calibrated to a global number, this equality holds by construction rather than serving as external validation. Further tests should compare the allocation with independent institutional records, evaluate sensitivity to dominant centers, and assess whether country ranks are stable across defensible alternatives. When ranks are unstable, ordinal claims of relative ownership should be withheld or qualified.

The main methodological criticism of a single number is that it invites different audiences to supply their own meaning. A statistical agency may read it as accounts coverage, a fiscal authority as a compliance gap and a commentator as inequality. These uses require different observational support. Publishing a compact estimate dictionary alongside any empirical results preserves the interpretation and supplies a basis for reproducibility: reference date, unit, currency, asset perimeter, residence basis, treatment of intermediary entities and debts, data vintage, correction assumptions, unallocated share, and uncertainty. These fields are ordinary requirements of transparent empirical measurement, not a proposed new theoretical system.

A Reproducible Research Workflow

An empirical implementation should start with a written target specification before downloading any data. The unit of observation can be a center-by-origin-by-instrument-by-period cell when sufficient public statistics exist, complemented by a separately governed administrative record system. The researcher must decide which economies belong to the center perimeter using documented criteria and preserve sensitivity to alternate definitions. A country with a large financial sector is not automatically a secrecy jurisdiction, and the use of a center list for estimation does not establish a legal or normative judgment.

The next stage constructs a same-vintage source archive. For each dataset, preserve the release date, table identifier, series keys, instrument, sector, currency, valuation and relevant statistical notes. The archive should preserve the original extract and the transformation script. Public series can be redistributed under applicable terms; sensitive records should stay within the authorized secure environment. In both settings, evidence ledgers should permit external inspection of the reasoning without exposing account identities.

A portfolio reconciliation aligns debtor and creditor sides of the international accounts and documents all known exclusions. Adjustments for missing reporting economies, institutional investors, government holdings and inconsistent instrument classifications should be traceable. A banking component then identifies relevant deposit liabilities and documents sector coverage. Conversion to broader asset stocks, where used, must state which parameters measure unrecorded coverage and which express instrument composition. Corrections used to reconcile different series must not be applied twice.

Ownership estimation follows rather than precedes this reconciliation. Construct immediate counterparty and ultimate resident-owner matrices separately when the data allow it. Specify how intermediate entity positions are reallocated, how corporate and nonprofit holdings are treated, and how unknown or ambiguous ownership is represented. Any imputed matrix requires a reason for the selected proxy, time variation and uncertainty. The purpose is to make decisions inspectable, not to manufacture complete cells.

Thereafter validate aggregate and country-level results, preserve alternative estimates, and compute ratios using harmonized definitions. Tax compliance or wealth inequality extensions should link only the amount that is missing from the relevant target records and account for overlapping sources. Published empirical tables should present benchmark and sensitivity results with status flags indicating observed, estimated, imputed or suppressed inputs. A reconciliation table should identify any discrepancy left unexplained.

For policy studies, preregister or document research choices, data revisions, treatment timing, outcome variables and identifying assumptions. Replication of a published study requires the original code, data access where allowed and an explicit definition of the replication target. An abstract or data availability statement does not verify an empirical result independently. This dossier reviews source claims and methodological conditions, but does not run such a replication. Its numerical illustrations are reproducible arithmetic, not research estimates.

Finally, publication schedules should distinguish the reference year from the release date and the revised version of the dataset. A new observation can require revisions to both the stock and the distribution. The public report should explain them. In particular, source changes that affect formerly missing owner populations deserve prominence because these changes can alter the interpretation of historical trends as well as their size.

Economic Interpretation and Research Priorities

The principal economic advantage of estimating offshore household wealth is that international accounts and distributional statistics can better represent who holds financial claims. A country whose residents own substantial unrecorded assets will have an understated foreign asset position, all else equal. Correcting that position can affect evaluations of national balance sheets, income accruals and external vulnerability. It does not necessarily imply that the assets can be mobilized by the government or domestic banking system during a crisis. Accessibility, ownership distribution, liquidity and legal restrictions are separate questions.

A resident-owned foreign asset also has different effects on a household's balance sheet and the domestic productive capital stock. A household that transfers a deposit and retains the same ownership may change the geography of intermediation without changing its gross wealth. A resident-owned foreign fund may invest in home-country securities; after tracing the network, the ultimate risk exposure can remain domestic despite a foreign intermediary. The economic effect of the transaction depends on its financing, use of funds and market counterparties. Offshore wealth is thus an informative ownership category but is insufficient as a standalone measure of domestic capital scarcity.

For fiscal research, a useful denominator is the potentially taxable asset or income base under applicable law, followed by estimated declaration and enforcement coverage. For inequality research, the relevant baseline is the particular wealth distribution and the missing amount after accounting for already included assets. For international financial accounts, the focus is residence, valuation and counterparties within the applicable statistical standard. The same empirical stock may enter all three research domains through different transformations that should remain explicit.

The literature also suggests that the most useful future data improvements are not confined to greater aggregate reporting. Harmonized immediate and ultimate counterparties would make the geographic map more accurate. Depositor-sector detail and reliable instrument composition would improve the attribution of financial wealth. Administrative linkage would help distinguish ownership from visibility. Asset-class bridges would reduce double counting. Better measures of coverage outside major centers would constrain sensitivity ranges. These priorities follow from existing source limitations and are offered as research recommendations rather than a new doctrine.

Policy evaluation should use multiple outcomes. A reduction in noncompliance is a legitimate success even if private portfolios remain globally diversified. A decrease in observed balances is ambiguous unless it can be assigned to transactions, coverage, revaluation or residence changes. A decline in measured wealth at one location may coexist with stable global offshore wealth. A statement about the success of transparency should identify which extensive or intensive margin is being measured: detection, declaration, tax assessment, location, asset class, or aggregate amount.

Economic interpretation also requires proportionality. Offshore household wealth has both legitimate and unlawful uses, varying by the individual and arrangement. Statistical results about the distribution of assets are not evidence of criminal responsibility in an individual case. Enforceable compliance obligations and adjudication must be established with the relevant law and facts. The role of aggregate measurement is to quantify financial relationships, improve the understanding of statistical omissions and support properly identified economic analyses.

Conclusion

Measurement of offshore household wealth is a problem of economic attribution across multiple layers of financial recordkeeping. The global portfolio discrepancy offers an important route to estimating missing securities; bank statistics provide complementary evidence about deposit claims and immediate counterparties; country allocation requires an explicit bridge to ultimate household residence; and administrative matching can illuminate declaration status and distribution. These tools answer distinct questions whose results become valuable together when their assumptions and scopes are kept visible.

The published literature establishes the material scale of household offshore financial assets under its respective definitions and demonstrates that the geography of their administration and ownership evolves. Historical estimates do not become live observations through repetition. Differences in wealth and GDP denominators, instrument composition, data vintage and coverage must be maintained in comparisons. Statistical noncoverage is not equivalent to tax evasion, and holdings legitimately declared to authorities may remain offshore.

The central analytical result of this dossier is that the distinction between ownership, statistical recognition and fiscal disclosure changes the interpretation of nearly every measurement method. The aggregate wealth stock is compatible with many allocations across countries and households; the same degree of statistical coverage is compatible with different compliance outcomes. Correcting international accounts, estimating inequality and evaluating transparency reforms therefore require different identified increments of the underlying financial positions.

A reliable research program preserves the target definition, reconstructs cross-border claims without duplicating intermediaries, accounts for valuation and residence changes, maintains unassigned and uncertain shares, and validates estimates with genuinely independent evidence. Where these steps cannot be performed, the uncertainty is part of the result. Improvements in information infrastructure should be evaluated in terms of the observational dimension they improve rather than assumed to eliminate every form of offshore activity. This is the basis for a disciplined scientific understanding of offshore household wealth: quantitatively informative, transparent about identification, and consistent with the economic meaning of ownership.

References

Alstadsæter, Annette, Matthew Collin, Bluebery Planterose, Gabriel Zucman, and Andreas Økland. 2025. Who Owns Offshore Real Estate? Evidence from Dubai. EU Tax Observatory Working Paper 1, revised September 16. Source.

Alstadsæter, Annette, Niels Johannesen, and Gabriel Zucman. 2018. Who Owns the Wealth in Tax Havens? Macro Evidence and Implications for Global Inequality. Journal of Public Economics 162: 89–100. DOI: 10.1016/j.jpubeco.2018.01.008. Source.

Alstadsæter, Annette, Niels Johannesen, and Gabriel Zucman. 2019. Tax Evasion and Inequality. American Economic Review 109 (6): 2073–2103. DOI: 10.1257/aer.20172043. Source.

Alstadsæter, Annette, Sarah Godar, Panayiotis Nicolaides, and Gabriel Zucman. 2023. Global Tax Evasion Report 2024. EU Tax Observatory. Source.

Bank for International Settlements. n.d. Locational Banking Statistics: Overview and Glossary. BIS Data Portal. Accessed November 18, 2026. Source.

Beck, Roland, Antonio Coppola, Angus J. Lewis, Matteo Maggiori, Martin Schmitz, and Jesse Schreger. 2024. The Geography of Capital Allocation in the Euro Area. ECB Working Paper Series 3007. DOI: 10.2866/3067767. ECB author profile with study abstract. Source.

Bénétrix, Agustín, Lorenz Emter, and Martin Schmitz. 2024. Automatic for the (Tax) People: Information Sharing and Cross-Border Investment in Tax Havens. Economic Policy 39 (120): 853–895. DOI: 10.1093/epolic/eiae041. Source.

Boas, Hjalte Fejerskov, Niels Johannesen, Claus Thustrup Kreiner, Lauge Truels Larsen, and Gabriel Zucman. 2024. Taxing Capital in a Globalized World: The Effects of Automatic Information Exchange. NBER Working Paper 32714. DOI: 10.3386/w32714. Research publication page. Source.

European Commission, Directorate-General for Taxation and Customs Union, and ECORYS. 2021. Monitoring the Amount of Wealth Hidden by Individuals in International Financial Centres and Impact of Recent Internationally Agreed Standards on Tax Transparency on the Fight Against Tax Evasion. Luxembourg: Publications Office of the European Union. DOI: 10.2778/647791 (PDF). Source.

Faye, Souleymane, Sarah Godar, Carolina Moura, and Gabriel Zucman. 2025. Global Offshore Wealth, 2001–2023. EU Tax Observatory Working Paper 41, December. Source.

International Monetary Fund. n.d. Portfolio Investment Positions by Counterpart Economy (PIP), formerly Coordinated Portfolio Investment Survey. IMF Data. Accessed November 18, 2026. Source.

Londoño-Vélez, Juliana, and Javier Ávila-Mahecha. 2021. Enforcing Wealth Taxes in the Developing World: Quasi-Experimental Evidence from Colombia. American Economic Review: Insights 3 (2): 131–148. DOI: 10.1257/aeri.20200319. Author-institution research summary. Source.

OECD. 2014. Recommendation of the Council on the Standard for Automatic Exchange of Financial Account Information in Tax Matters, OECD/LEGAL/0407, Annex I, Common Standard on Reporting and Due Diligence for Financial Account Information. Legal Instruments database, accessed November 18, 2026. Source.

OECD. 2025. Consolidated Text of the Common Reporting Standard (2025): Standard for Automatic Exchange of Financial Account Information in Tax Matters. Paris: OECD Publishing. DOI: 10.1787/055664b1-en. Source.

Zucman, Gabriel. 2013. The Missing Wealth of Nations: Are Europe and the U.S. Net Debtors or Net Creditors? Quarterly Journal of Economics 128 (3): 1321–1364. DOI: 10.1093/qje/qjt012. Source.