RESEARCH DOSSIER
Offshore Securities Holdings
Abstract. Offshore securities holdings connect legal issuers, institutional and household investors, financial centers, custodians, funds, corporate groups, and underlying markets across jurisdictional boundaries. This Research Dossier establishes the economic and statistical distinctions necessary to interpret these connections. It examines portfolio investment coverage, securities custody, issuer residence and parent nationality, alternative measures of operating exposure, fund look-through, valuation and stock-flow reconciliation, international data gaps, historical empirical evidence, household wealth, financial stability, and transparency policy. Using a comparative review of official methodologies and original research, together with explicitly hypothetical accounting examples, it demonstrates how geographical reallocation can preserve a portfolio's aggregate value while materially changing bilateral exposures, and how multiple balance-sheet layers can produce double counting if consolidation is undefined. The study distinguishes published empirical findings from analytical inference and identifies the limits of observed and estimated holdings. Its conclusion establishes a purpose-specific approach to complementary residence-based, consolidated and ultimate-exposure statistics, with documented coverage, uncertainty, and valuation.
Keywords: Offshore economics; securities holdings; portfolio investment; financial centers; custody; issuer residence; corporate nationality; beneficial ownership; investment funds; financial statistics; valuation; financial stability
Introduction
A cross-border security can be issued by a company incorporated in a jurisdiction in which it employs no significant workforce, held by a pension fund resident elsewhere, administered by an intermediary in a third jurisdiction, denominated in a fourth country's currency, and ultimately financed by households in several economies. Each connection has a genuine economic meaning. The problem of offshore securities holdings arises when information about one of those connections is used as evidence about another.
This monograph studies the economic location of financial claims embodied in transferable instruments. Its central finding is that offshore securities holdings constitute several distinct layers of international portfolio investment: the residence of the immediate counterpart, the consolidated exposure of the issuer group, the distribution of underlying assets, and the location and sector of the ultimate investor. A defensible measurement strategy preserves the original positions while reconstructing specific alternative allocations for specific questions. The same value cannot be counted as several independent units of household wealth simply because it passes through several corporate balance sheets.
The research belongs to Offshore Economics. It concentrates on positions, their allocation, valuation, and implications for economic incidence and risk. The economic organization of the investment management industry, the estimation of total offshore household wealth, the accounting of foreign direct investment, illicit-financial-flow measurement, and substantive securities law are distinct research objects. They enter here where they affect the interpretation of securities positions. Offshore status alone says nothing conclusive about legality, tax compliance, or investor intention.
The evidence combines international statistical definitions, national statistical practice, historical empirical work, and contemporary institutional analysis. Numerical examples constructed in the text are explicitly hypothetical and use arbitrary monetary units. Published estimates are identified by period and population. No newly downloaded security-level holdings matrix or replication of a published empirical estimate is presented. The contribution is an integrated economic interpretation and a set of transparent accounting demonstrations that connect the source literature with the accuracy requirements of policy analysis.
The economic object and the units of observation
The relevant object is a financial position in an equity security, investment fund share, or debt security at a specified valuation date. The investor has a claim on an issuer; the corresponding issuer records an obligation or equity interest. Such an instrument can be transferred, and its price can change independently of the initial cash proceeds that financed the issuer. Portfolio investment is narrower than all securities holdings. Some securities belong to direct investment relationships or reserve assets, and derivatives are a separate instrument class. A research universe must distinguish a broad securities dataset from the statistical functional category it is intended to reproduce.
The IMF's portfolio survey scope (2018) describes assignment by issuer residence and allows aggregate or instrument-level data collection. Its current Portfolio Investment Positions dataset separately identifies equity and fund shares, longer-term debt, and shorter-term debt. These categories determine what the observation represents. A bank loan, deposit, derivative, privately held nonsecurity claim, and securities portfolio are economically connected but cannot be treated as interchangeable stocks.
Consider four positions of equal market value. A household owns shares in a foreign operating company through a domestic broker. Another household owns units in a foreign fund whose investment is in domestic companies. A bank holds bonds issued by the offshore financing subsidiary of a multinational. A domestic institution holds government bonds in an account at a foreign custodian. All four positions can involve an offshore element in a suitable geographical comparison. Only some are claims on nonresident issuers for the particular domestic reporting economy. The first has no necessary offshore financial-center involvement; the second introduces an interposed fund; the third changes the institutional geography of corporate financing; the fourth changes the location of securities administration without necessarily creating a foreign portfolio asset.
The analytical unit should therefore contain at least the instrument, position date, value and valuation basis, holder institutional sector and residence, issuer residence, and whatever independently supported parent, custodian, or underlying exposure information is available. Data on the natural person ultimately interested in a security will often be unavailable. In that event an unknown category preserves the distinction between information absent and information demonstrating domestic or foreign ownership.
The adjective offshore is relational. A location may provide international capital-market services to nonresidents while maintaining a substantial domestic economy. An empirical study must specify the reference economy and selected financial-center universe, including the criteria for inclusion. A corporation's legal domicile, a financial center's service specialization, and a particular account's classification are different properties. Country lists can support a reproducible tabulation, but geography alone cannot identify a legal violation or a tax objective.
The institutional chain of ownership and custody
A security holding commonly involves execution, settlement, recordkeeping, custody, and asset management. These functions can be performed by different institutions. The broker executes transactions; the custodian maintains assets and records; the asset manager makes delegated investment decisions; a fund or other vehicle owns assets on its balance sheet; the ultimate household or institution may own claims on the vehicle. The contractual allocation of responsibility and the statistical allocation of a position are related through the structure of these relationships.
An omnibus account illustrates the information problem. A local custodian may observe a foreign intermediary as its immediate customer, and that intermediary may maintain records allocating the aggregate holdings among its clients. A researcher with access only to the local custodian's records observes a real cross-border service relationship, but lacks the investor geography. If the total in the omnibus account is 100, a defensible inference allocates 100 to the intermediary location only in a dataset explicitly describing that observed relationship. The equivalent 100 in the investor residence distribution may be unknown. A measurement system that allocates both observations as independent holdings creates a false increase in the economic total.
The U.S. Treasury's TIC methodological FAQ (n.d., question 7) documents such distortions in country reporting of foreign-owned U.S. securities and distinguishes them from the geographical recording of securities purchases. The geographical location of the transacting intermediary may differ from that of the investor who acquires exposure. Consequently, tracing flows by the geography of the trading desk is inappropriate when the research objective is accumulation by resident holders.
There is also a distinction between a custodian holding an instrument for a customer and a financial vehicle holding an instrument for its own account while issuing a further claim. In the first case, the custody chain generally represents successive records concerning the same asset. In the second, there are two different financial assets: the final investor's share or debt claim against the vehicle and the vehicle's claim on the underlying issuer. The institutional balance-sheet approach retains both genuine contracts. A consolidated exposure approach eliminates the claim internal to its chosen perimeter. The perimeter is a substantive choice, because it determines whose balance sheet is being analyzed.
The informational consequences are asymmetric. A holder survey may identify the instrument and resident investor while the issuer-side records identify a foreign custodian. An issue database can identify the instrument and issuer group without recording its current holders. A tax administration may observe an account holder or controlling person with information restrictions that prevent direct statistical use. Record linkage can improve coverage when these sources are legally and technically compatible; treating one source as a complete substitute for another generally reduces interpretability.
This leads to an important distinction between ignorance and misallocation. Misallocation occurs where an observed item has been assigned to a country inconsistent with an expressly chosen classification. Ignorance occurs where the variables needed to implement that classification are absent. Both may produce biased aggregates, but only the first gives an observed basis for deterministic reassignment. The second requires estimation, bounds, or an unresolved category.
Why securities are issued and held offshore
Offshore issuance is part of the institutional organization of financial markets. Potential benefits include access to a larger investor pool, standardized documentation, currency funding, specialization in asset servicing, and the ability to pool risks or isolate particular contracts. Costs include administrative complexity, information asymmetry, additional counterparties, legal fragmentation, and dependence on the continuing operation of service providers and regulatory permissions. The value of an offshore arrangement depends on its economic function and the institutional constraints of the parties.
A financing subsidiary illustrates the allocation of functions. Suppose a parent corporation operates in country A. A subsidiary in center B borrows from market investors and advances the proceeds to affiliated entities. Investors purchase the subsidiary's debt. The group obtains funding even though its consolidated productive activity may be elsewhere. The parent may support the subsidiary through a guarantee, an internal contractual claim, or expectations of group support; these arrangements imply different legal and financial exposures. Economically, a financing subsidiary can separate the market-facing funding contract from the operational allocation of cash. This explanation does not establish why any specific company selected its domicile.
Coppola, Maggiori, Neiman, and Schreger (2021), in their global capital-flow remapping, associate offshore-issued traded securities with parent entities to examine how the geographic picture of international finance changes. The logic is central to the present inquiry: the observed place of issuance and the inferred location of a consolidated corporate borrower answer different questions. Classification by nationality typically changes the geographical allocation more readily than the overall quantity of outstanding debt.
For investors, pooled vehicles may provide diversification, operational economies of scale, institutional access to markets, and professionally delegated portfolio administration. Uniform fund domiciles can reduce duplication of administrative structures across markets. Their specialization may also concentrate economic service rents in relatively small jurisdictions. High financial assets relative to local GDP can emerge from this specialization without implying that local households own the assets.
Contracting frictions matter. Each intermediary needs information to verify entitlements, supervise delegated operations, and manage legal and operational risk. A structure that reduces transaction costs for an investor can impose costs on statisticians, tax authorities, and financial supervisors. These costs are external to the immediate contractual price unless institutions internalize them via reporting, audit requirements, or cooperation. Conversely, overly duplicative reporting consumes resources and may reduce access to international markets. The public objective is therefore proportionate transparency: information sufficient for the relevant collective purpose, with clear safeguards for legitimate confidentiality.
Tax differentials may influence both issuance and portfolio domiciliation, especially where withholding, entity taxation, or the treatment of distributions affects after-tax returns. Yet the observed use of a particular location cannot isolate this effect from investor access, language, legal infrastructure, scale, regulation, and historical specialization. A credible claim about the quantitative influence of tax requires a research design that controls or otherwise accounts for these competing explanations.
Residence, parent nationality, and ultimate exposure
The balance-of-payments framework organizes economic relationships by residence. The IMF balance-of-payments dataset defines its object as transactions during a period between residents and nonresidents. Positions at a date are a complementary stock concept. Residence allocation provides a consistent cross-jurisdiction accounting boundary essential for national balance sheets, external financing and the measurement of claims between institutional units.
The issuer's nationality in this literature is an attribution to a corporate parent or economically meaningful group identity, not the citizenship of a natural person. It may depend on industry conventions and the information available about head offices, management or operations. A nationality map of corporate debt is different from a map of the legal obligors; a map of operating exposure may differ from both. For an asset-backed security, the distribution of the collateral may be more informative for credit risk than the domicile of the issuing vehicle.
The BIS international debt securities methodology distinguishes issuer residence and parent headquarters and reports international issuance on a face-value basis. It also distinguishes issuance outside the local market from domestic issuance using market-related characteristics. These distinctions mean that international issuance is neither a synonym for offshore issuance nor a direct observation of investments by foreign holders. A resident issuer can issue international securities purchased domestically, and a foreign investor can purchase domestic-market debt.
Suppose an offshore subsidiary B has 100 of bonds held by investors in C. Parent A guarantees 100 of principal. The residence matrix contains a claim from C to B. A parent-exposure matrix places that claim from C to A. An enforceable guarantee may create a contingent obligation of A, which must be considered in a default scenario; the guarantee does not make the investor's bond two simultaneous assets of 100. If the subsidiary has independent revenue or financing rights, attributing all expected recovery to the parent alone can also misstate the exposure.
The question of ultimate exposure is especially challenging for equities. A multinational's aggregate income depends on sales, costs, taxes, financial commitments and economic conditions across many countries. Corporate headquarters is useful for governance and some financing questions but can be a weak proxy for the location of future cash flows. A revenue-based mapping may improve an analysis of demand exposure, while a production-based map can illuminate supply chains or labor costs. Neither gives a sufficient measure of sensitivity to an exchange-rate movement or a geopolitical shock without additional information.
Bertaut, Bressler, and Curcuru (2020) examine this issue in their revenue-based portfolio analysis. Their estimates illustrate why the nationality correction need not exhaust the economic reallocation. They use country revenue shares and approximate regional allocations, making the assumptions transparent. The resulting map is an exposure proxy, not the country's ownership of the security or a complete measure of asset return covariance.
An appropriate dataset maintains a residence column and preserves the original observed value while adding a dated, documented exposure map. Different maps can then be compared. A single supposedly corrected nationality code without a record of the evidence and purpose of the change makes it impossible to distinguish corporate restructuring, methodological revisions, and actual changes in international investment.
International statistical sources and coverage
The coordinated portfolio survey, now published as PIP, is the principal cross-country framework for allocating reported cross-border portfolio assets to counterpart economies. Its voluntary reporting nature matters when constructing a global matrix. A liability allocation derived from partners' asset reports is useful when debtor-side geography is incomplete, but it inherits the reporting coverage of the creditor side. A zero in a published extract must be interpreted with the associated reporting metadata: it may represent an observed zero, suppressed information, a missing reporter, or a classification that does not apply.
The Handbook on Securities Statistics, prepared jointly by the BIS, ECB, and IMF (2015), provides a common methodological basis for securities classifications and valuation and for distinguishing issue and holdings tables. That framework aids the integration of datasets but does not make datasets with unequal populations interchangeable. The IMF overview also indicates a future alignment of the handbook with the updated national-accounts framework. A time series must therefore record the standards actually used by compilers rather than assume simultaneous adoption of newer manuals.
BIS international debt securities give the issuance side for a defined class of traded debt instruments. They discriminate among the issuer's residence, nationality, currency and maturity, but the face value of the instrument is not its observed market value in the investor's portfolio. A position matrix requires holder data in addition to issue data. The currency of denomination must also remain independent of the geography of either party. An obligation denominated in dollars does not necessarily have a U.S. issuer or a U.S. holder.
The ECB's securities holdings methodology (n.d.) describes an instrument-level approach and identifies holder sector and country as dimensions of euro area data. Linking positions with reference data is valuable because the detailed characteristics of a financial instrument change less often than its holder composition. The approach also makes duplicate records easier to diagnose, although confidentiality and incomplete coverage limit access to a comprehensive international matrix.
The U.S. Treasury International Capital system provides a national example of published holdings and the geographical problems in allocating them. Its asset and liability surveys need not share the same measurement bias. An instrument identifier can reveal issuer location on an asset survey even if a foreign custodian's records do not identify the final holder on the liability side. This asymmetry is important for cross-country reconciliation.
For long-period international comparisons, Lane and Milesi-Ferretti (2017) assemble an external positions dataset covering more than 210 economies over 1970–2015. The historical work is an example of the value of integrating national and international statistics over time. Its scope is broad external investment, not a direct register of offshore securities ownership.
A reproducible research file should identify the dataset edition, period, reporting population, currency and valuation units, functional category, sector and instrument, residence convention, treatment of missing observations, and relevant transformations. Combining two publications without these descriptors can yield an apparent discrepancy that disappears once their universes are harmonized. An international securities statistic is scientifically useful precisely when its construction is understood well enough to establish which inference it can support.
Stocks, transactions, prices, and exchange rates
A securities holding is valued at a point in time, whereas acquisitions and disposals occur over a period. Let V0 and V1 denote comparable beginning and ending positions. Let T denote net transactions; let P and X denote price and exchange-rate revaluations under a consistent decomposition; let O denote other volume changes and reclassifications. A standard reconciliation is V1 = V0 + T + P + X + O. Any remaining statistical discrepancy should be identified separately where the data do not close.
For a simple hypothetical account, the opening stock is 100. Net acquisition is 15, security price appreciation contributes 8, currency translation reduces the reported value by 5, and a classification change removes 3 from the selected universe. The closing stock is 115. Its growth of 15 equals the quantity of transactions only by coincidence: the other effects happen to sum to zero. Observing the opening and closing balances alone would not identify the individual components.
A bond's market price can fall when the yield rises while the nominal obligation remains unchanged. An equity portfolio can become larger in dollar terms despite net sales if share prices rise sharply. If one country's currency appreciates against the reporting currency, the translated value of securities denominated in that currency may rise without any change to the number of instruments held. Currency and price effects interact when exposure is measured over a period; the chosen decomposition must state how cross-products are allocated.
Consider a bond with face amount 100 initially priced at par. It now trades at 90. A holder-position survey at current market value records 90 while a face-value issue table records 100. The difference of 10 is not an unidentified foreign asset, a missing investor, or evidence of fiscal evasion. It is an expected result of comparing different valuation bases.
Cash flows must also distinguish replacement funding from the net increase in obligations. A company refinancing 100 of maturing securities by selling new bonds of 100 has gross issuance and gross redemption equal to 100, but no increase in outstanding principal in this closed example. The investor allocation may change completely from one country to another. Consequently, matched aggregate issue volumes do not establish unchanged exposure networks.
An empirical analysis of offshore financial growth should compare changes in positions after measuring these effects. A surge in a financial-center position could reflect equity valuation, a change of fund domicile, an issuer merger, improved survey participation, or increased investment. Different causal interpretations imply different economic mechanisms. Without a reconciliation, the defensible description is growth of the recorded stock, not an asserted increase in capital flight or newly raised financing.
Bilateral matrices and the missing-owner problem
Let H(i,j) be an observed portfolio asset of investors resident in economy i against immediate issuers resident in economy j. For an instrument-specific matrix with harmonized valuation, row sums describe assets held by participating creditor economies; column sums approximate liabilities to those creditors. A residence matrix is useful even in the presence of offshore intermediation because it describes institutional claims associated with those reporting boundaries.
If a security held by a resident investor is omitted from the survey, the actual asset may still be included in the debtor's reported liabilities. Define a comparable aggregate discrepancy D as reported liabilities minus reported assets for a matched perimeter. A positive D may be consistent with omitted assets. It can also arise from overreported liabilities, inconsistent prices or exchange rates, incomplete sampling, reporting dates, differences in the treatment of securities lending, or unmatched instruments. The discrepancy is an identification problem rather than an estimate of concealed wealth by definition.
Take an illustrative closed three-economy example in which issuers report 300 of liabilities to foreigners. Matched creditor positions sum to 270, giving a gap of 30. Suppose 10 is identified as a valuation mismatch and 5 as a known population mismatch in a reconciled version of the accounts. The remaining comparable net gap is 15. It may contain omitted assets, other data errors, or both. It cannot identify the investor country by itself. Independent evidence about investor residence and account coverage is required before allocating the residual.
A mirror statistic is a partner's reported counterpart to a missing or uncertain national position. It is strongest when both datasets use the same perimeter, date and value basis and when the partner's view correctly identifies the geographic counterparty. A mirror based on residence of a custodial intermediary may correct the size of a sector aggregate but not the distribution of end investors. Mirror allocations must be marked as estimates and their uncertainty presented separately from direct observations.
A useful sensitivity exercise can be performed without pretending that missing positions are observed. Suppose total known positions in an offshore custodian account are 100, of which 70 can be assigned with evidence to investor economies and 30 remain geographically unresolved. If a particular investor country has 20 of the known positions, the simplest nonnegative bound for that country's total under this closed universe is 20 to 50. Additional credible restrictions on the unallocated 30 can narrow the interval. Reporting a point estimate of 35 requires an explicit allocation assumption, and the midpoint is not intrinsically more reliable simply because it is centered.
This logic also applies to data suppression. Analysts should separate confidential but included aggregate positions, wholly absent reports, and unresolved geographical codes. The first may allow a total to be estimated accurately while withholding details; the second undermines global completeness; the third primarily changes allocation. Each has a different effect on policy inference.
Security-level reallocation and consolidation
Security-level data allow a researcher to map each instrument to alternative counterparties while retaining its original position. Let h(i,s) denote a holding by investor economy i of security s. Let w(s,k) be a nonnegative allocation weight for an exposure economy k, including an explicit unknown economy. When the weights for each security sum to one, the alternative exposure for investor i in k is A(i,k) = sum over s of h(i,s) times w(s,k). For a fixed instrument population, this transformation preserves the aggregate position across k. An alternative valuation, leverage adjustment, or derivative sensitivity would require a separate reconciliation and may not preserve the same total.
For a hypothetical investor portfolio, 60 is held in the debt of an offshore financing subsidiary, 40 in the equity of a corporation resident in an offshore center, and 100 in directly held government bonds. The original three holdings total 200. If the first security is allocated entirely to its documented parent in economy A, and the second to corporate exposure countries B and C with weights 0.75 and 0.25, the reallocated positions are 60 in A, 30 in B, 10 in C, and 100 in the government bond jurisdiction. They still total 200. The residence map and the exposure map are two views of one portfolio.
Fund shares require additional care. Assume a resident household holds 100 of units in an offshore fund. The unleveraged fund holds underlying securities worth 100, consisting of 60 of bonds in country A and 40 of equities in country B, with negligible other assets and liabilities. The household's direct contractual asset is the fund share worth 100. Its mechanical portfolio exposure through the fund totals 100: 60 and 40 allocated to the underlying securities. Counting both maps as external household assets valued at 200 doubles the position. At the level of unconsolidated institutional balance sheets, however, the household's claim and the fund's separate underlying assets are valid observations.
Now assume the fund borrows 50 and holds securities worth 150. Net asset value is 100 before fees and other liabilities. If the fund holds 90 of securities in A and 60 in B, its gross underlying holdings sum to 150. A simple allocation of the household's net 100 based on asset shares gives 60 and 40, while gross exposure to the assets is 90 and 60. Neither representation captures the full effect of the fund's borrowing on household risk. The liability of 50 must be tracked to understand leverage, loss amplification, and refinancing. The gross exposure and equity-at-risk measures should carry distinct labels.
In layered funds, blindly following investments recursively can create loops, double counting and unexplained residuals. A look-through computation should define a terminal set of instruments, track ownership weights and the time at which they are observed, and stop at unresolved structures when circular investments or insufficient information prevent reliable attribution. It should also distinguish market weights from voting control, because ownership concentration for governance purposes can differ from exposure concentration.
The primary scientific benefit of security-level reassignment is traceability. A researcher can identify whether a changed bilateral total comes from new instrument weights, a new parent linkage, changing portfolio composition or updated valuation. A country-level correction factor may be useful when microdata are inaccessible, but applying a factor from one instrument class or period to another assumes stability that needs independent support. Validation requires a matched subset, coverage reporting, alternative classification tests and monitoring of reassignment residuals.
Empirical findings on offshore issuance and portfolio geography
Several sources establish that offshore structures can materially affect measured investment geography. Their estimates address different populations and should be kept separate. Bertaut, Bressler, and Curcuru (2019), in Globalization and the Geography of Capital Flows, analyze the U.S. cross-border portfolio and report about $12 trillion of securities holdings at the end of 2017. Their residence-based common-stock positions in the selected offshore and low-tax centers rose from approximately $400 billion in 2003 to more than $2.2 trillion in 2017. They use different remapping evidence for equities, corporate debt, asset-backed instruments and fund shares.
The estimates have a specific source universe: U.S. portfolio claim surveys, instrument reference data, and allocations for particular investment vehicles. The reported financial-center total is neither a global offshore wealth estimate nor a measure of tax evasion. The common-stock series cannot be directly equated to the corporate-bond series. The chosen list of centers is a research classification, not a legal finding concerning those jurisdictions.
The 2021 study by Coppola and colleagues reports that assigning offshore-issued securities to their corporate parents reveals nearly $600 billion of additional U.S. positions in Chinese firms in the authors' historical case. This figure is a result of an alternative geographical allocation, not a later revision independently calculated for this monograph. The result illustrates that an investment classified against a financial center can finance firms with economic connections to large emerging markets. It does not establish that every exposure reassigned to Chinese firms is a direct claim on a Chinese-resident legal entity.
A further distinction concerns multinational activity. In the 2020 Federal Reserve note, the authors examine roughly $27 trillion of U.S. investor equity holdings at the end of 2018, combining domestic and foreign holdings. Their estimate of domestic exposure declines from about $21.8 trillion on a nationality basis to about $18 trillion when they allocate equities by the geography of corporate revenue. Their data and regional imputation are specific to the study period. Revenue allocation is especially relevant for the distinction between apparent portfolio home bias and exposure to foreign final demand.
These results are mutually informative without being additive. The earlier note studies U.S. external holdings; the later note includes domestic equities and uses another exposure measure; the Coppola study uses its own security and parent mapping. Adding the three reallocation magnitudes would sum overlapping exposures, different periods, and different denominators. The appropriate comparison instead evaluates what direction and magnitude a particular change of classification produces within its own population.
The broader economic interpretation is that financial globalization can be underestimated or overestimated depending on the metric. A residence-based measure may overstate a country's investment in a financial center and understate its exposure to operating firms elsewhere; a consolidated measure can then decrease one international link while increasing another. Revenue mapping may reveal international exposures within nominally domestic portfolios. Neither transformation makes cross-border financial contracts disappear from the legal balance-sheet record. Each identifies a distinct economic layer.
An official national measurement example
The U.S. final June 2024 holdings survey (Treasury, 2025, Table A) records foreign portfolio holdings of U.S. securities with a total of $30,881 billion. It reports $16,878 billion in equities, $12,688 billion in long-term debt, and $1,314 billion in short-term debt. The three published components sum to $30,880 billion; the difference of $1 billion from the published total is compatible with independent rounding. The official total should be retained rather than silently changed.
These values are the national survey's measured foreign holdings of U.S. instruments, not foreign holdings through offshore centers alone. They include more than one institutional sector, and the allocation to a particular foreign economy can reflect the custody limitation discussed above. The survey's national aggregate is generally informative about the total claims in its measurement universe, even where the exact foreign owner country is less certain. The distribution must be interpreted using the method notes.
For this case the approximate proportion represented by equities is 54.7 percent (16,878 divided by 30,881), while long-term debt is about 41.1 percent and short-term debt about 4.3 percent. Individual rounded shares sum to 100.1 percent. These computations are descriptive calculations from the official 2024 data, not new empirical estimates of offshore holdings. The 2023 comparison is deliberately omitted because the release's narrative and Table A conflict on the earlier total; resolving that discrepancy would require consultation of the relevant benchmark report.
The case demonstrates three quality-control obligations. First, a cited data table must match the instrument universe of the assertion. Second, independent rounding should be acknowledged rather than treated as concealed data. Third, when a publication includes inconsistent narrative and tabular entries, the conflict must be retained in the evidence log and avoided in quantitative inference until resolved.
A further comparison with an issuer-level dataset would require selecting matching securities and harmonizing market and face values. Comparing the entire Treasury aggregate with BIS international issuance or with PIP assets for one jurisdiction would be inappropriate. The scope differs on both the instrument and holder sides.
The most policy-relevant lesson is that the reliability of a total and the reliability of its allocation can diverge. Misallocating 10 of an observed portfolio between two foreign countries leaves the global total invariant, whereas an omitted holding changes the total. Public commentary on the identity of a large creditor should therefore be more cautious than commentary on the aggregate amount whenever the data collection mechanism is vulnerable to custodial bias.
Currency denomination and the structure of financial risk
Currency provides a separate dimension of the geography of securities investment. A bond issued by an offshore subsidiary may be payable in U.S. dollars irrespective of the parent company's nationality, location of investors, or destination of financing. Currency exposure may reside partly with the subsidiary, parent, end borrower, or holder, depending on the currency of underlying assets, cash flows, and hedging contracts. An analysis of exchange-rate risk must preserve the denomination independently of the issuer classification.
Maggiori, Neiman, and Schreger (2020) document in International Currencies and Capital Allocation a strong preference among investors for debt in their own currency in the security-level dataset studied, with internationally held US dollar debt an important exception. The finding supports the importance of joint information on currency and geographical allocations. It does not imply that every foreign-currency bond is risky to a particular investor, because hedges, currency-linked revenues, liabilities, and investment objectives can alter effective exposure.
Consider a borrower with revenues of 100 in currency A and a debt obligation of 100 in currency B. If currency B appreciates relative to A, the local-currency cost of servicing the debt increases, absent offsetting hedges or revenue. Offshore issuance may make the contractual location of the borrowing different from the location of the cash flows that support it. For a consolidated analyst, parent and revenue location help identify vulnerabilities. For a liquidity manager, the currency and maturity schedule are essential.
A holder of the same bond may be exposed to changes in its price because of credit spreads, interest rates, currency translation, or liquidity. A currency-hedged investor can still suffer a substantial decline in price when credit risk increases. A currency hedge may also create margin requirements and dependence on derivatives counterparties. Conventional portfolio securities positions do not describe all of these derivative exposures.
The economic consequence is that financial center holdings should not be treated as a single homogeneous risk class. A short-term government bill, a leveraged corporate credit instrument, and an equity claim in a globally diversified fund have different payoff and liquidity properties. Classifying them together by vehicle domicile can obscure correlated credit, duration, and currency risks. Classifying them solely by currency can obscure common exposure to a parent group or collateral pool. A credible stress analysis therefore retains both instrument and network attributes.
A comparative scenario can examine higher international interest rates, a depreciation in the operating economy, or heightened redemption demand in funds. The relevant loss channels differ. The first affects duration and refinancing, the second changes cash-flow coverage, and the third influences market liquidity and funding. No numerical shock impact can be claimed without instrument-specific valuation and balance-sheet data.
Household wealth, missing assets, and distribution
Household securities holdings through offshore intermediaries are a subset of the broader system. They can generate measurement errors when households' assets are omitted from domestic statistical surveys, when nominee structures obscure sector assignment, or when investors hold shares in foreign entities whose underlying assets are not apparent in the household accounts. The problem can affect private wealth estimates, investment-income measures, and the analysis of inequality.
Zucman (2013), in The Missing Wealth of Nations, uses offshore information and international portfolio discrepancies to estimate that approximately 8 percent of global household financial wealth was held in tax havens in the historical setting of the paper, with much of it absent from official records. That is an estimated household-wealth share under the author's methods and historical data, not the fraction of contemporary global securities held offshore. Extending the figure to all institutional investors or treating it as a current estimate would change both denominator and period.
Alstadsæter, Johannesen, and Zucman (2018) extend the research agenda on who owns offshore wealth in Who Owns the Wealth in Tax Havens?. Their historical benchmark places household offshore financial wealth at approximately 10% of world GDP in 2007, with geographic allocation informed by bilateral banking data. This is a wealth estimate relative to GDP, rather than a securities-only total or current observation. Assigning offshore assets to households requires geographically specific assumptions and evidence. The present assessment does not replicate their estimates.
A household's financial wealth can include deposits, fund shares, directly held securities, and claims on other entities; corresponding debts reduce net wealth. A gross total of portfolio assets held by financial-center institutions measures a different object. For example, an offshore fund with assets of 100 may have investors from households, corporations and institutional funds. Allocating the entire 100 to personally owned offshore wealth is unjustified without investor-sector information and a consolidation rule.
Similarly, a household can hold a security legally through a foreign institution while reporting it for tax and statistical purposes. An unobserved position may result from a survey gap despite lawful reporting to another authority. A recorded foreign custody relationship is not evidence that a person failed to declare income. Statistical omissions and tax noncompliance overlap only to the extent supported by independent information about the reporting obligations and the actual conduct.
The consequences for inequality analysis depend on which households own the omitted assets, not only on their aggregate value. Assigning a missing amount proportionately to all households preserves the shape of a distribution under some assumptions; allocating it predominantly to the top changes concentration substantially. Neither allocation is identified by an aggregate financial discrepancy. Distributional analysis requires a separate bridge between offshore positions and household-level information, with appropriate protections for confidential records and uncertainty about nonobserved investors.
Comparative financial geography and specialization
Financial centers perform different functions in the institutional chain. Two jurisdictions showing similar levels of recorded foreign securities assets may specialize in different activities. One may host custodial service providers holding third-party assets; another may domicile investment funds that own portfolios; a third may be a major location for corporate borrowing through affiliates. Their measured external assets and liabilities reflect these distinct roles.
Historical U.S. portfolio research provides concrete examples. The analysis by Bertaut and colleagues (2019) uses information on vehicles in the Cayman Islands, Ireland and Luxembourg for selected fund-share allocations and uses different sources for corporate equity and debt. The Treasury FAQ also identifies examples of important custody locations in its description of country attribution. These findings are dated and function-specific; they do not justify a universal ranking of jurisdictions by secrecy or illicit activity.
A comparative design should begin with instrument composition and institutional sectors. Fund intermediation typically creates relationships between investors and the fund and between the fund and underlying issuers. Corporate finance vehicles create market liabilities and affiliated or other assets. Custody may generate limited balance-sheet exposure to the securities themselves while the institution holds contractual operational duties and earns service revenues. In national accounts, a large volume of client securities does not necessarily represent an equal volume of the custodian's own assets.
Ratios to domestic GDP provide a measure of financial positions relative to local economic output, but a large ratio is sensitive to the size of the denominator and the institutional mix. A small jurisdiction with a specialized global service sector can have very large external positions relative to GDP. This indicates financial intensity rather than an estimate of domestic residents' wealth. Its vulnerability to regulation or market shifts should be assessed through local employment, fee revenue, taxable activity, fiscal dependence, and actual liabilities, not merely the portfolio ratio.
For the same reason, comparing a fund center's total positions with a major economy's national household wealth requires a bridge across sectors and consolidation boundaries. Otherwise the comparison describes different levels of an ownership chain. Transparent comparison should set out the reporting jurisdiction, instrument type, holder sector, issuer geography, and whether the observations are client assets under administration or assets of institutional units.
Geographical categories should be revised when the institutional functions or available evidence change. A historical classification is useful for a longitudinal study precisely because it is explicitly fixed or revised using documented criteria. Retroactively altering the list to obtain a stronger result introduces selection bias. Classification should be performed before interpreting the outcome whenever feasible.
Liquidity, leverage, concentration, and contagion
Offshore securities holdings link markets through common portfolios, collateral, funding relationships, and asset-management channels. A disruption need not originate in the jurisdiction where claims are formally recorded. A borrower in an operating economy may experience deteriorating earnings, reducing the value of debt issued by an offshore affiliate; fund investors may redeem shares in another economy; intermediaries may sell securities in a third market to meet obligations.
The FSB's 2025 nonbank monitoring report covers financial intermediation across 29 jurisdictions, using data for 2024, and identifies bank and nonbank links through funding, lending and holdings of bank-issued securities. Its reported nonbank share of aggregate financial assets is 51 percent in its monitored framework. That broad sector share is not the share of assets held offshore, nor the fraction held in investment funds. The report provides institutional support for examining channels between sectors rather than treating their risks as independent.
In a hypothetical case, two funds each hold the same security valued at 50, and each also holds unrelated assets. If one fund sells its position rapidly at a lower price, the second may face a mark-to-market decline even if its own issuer continues to pay. The interconnection is through asset prices and common holdings. If the funds are leveraged and collateral terms are sensitive to valuation, additional asset sales may be needed to restore funding conditions. Custody and settlement disruptions constitute another possible channel, distinct from the credit risk of the issuer.
A simple balance-sheet example makes leverage explicit. A fund begins with assets of 150 and fixed borrowing of 50, so equity is 100. A decline of 10 percent in all assets reduces them to 135 and equity to 85; under the assumed fixed debt valuation and no hedging or other changes, the decline in equity is 15 percent. The illustrative exercise demonstrates why net asset holdings, gross market exposures and liabilities must be separately represented. It is not a forecast.
Concentration also changes with the classification. On a residence map, holdings in three separately domiciled financing vehicles may appear diversified across jurisdictions, yet the three vehicles may belong to one corporate group. On a group map, the common credit concentration becomes visible. Conversely, aggregation by parent can hide different collateral pools or legal protections among instruments. Risk measurement requires a structure that retains the individual contracts and consolidates only at a clearly designated perimeter.
Stress testing should combine exposures by issuer group and sector with currency, maturity, collateral, liquidity, counterparty support, funding terms and information about common holdings. Data on beneficial investors can help assess correlated redemption behavior if appropriately available, but the country of their custodian alone is insufficient. This distinction is crucial for cross-border supervision, because responsibility for a regulated vehicle and the economic origin of shocks may reside in different places.
Transparency, taxation, and the measurement-policy boundary
Information policy has several distinct purposes: national statistics, prudential supervision, market integrity, tax administration, and the prevention of financial crime. These purposes can benefit from common identifiers and compatible data collection while remaining governed by distinct legal powers, access controls and confidentiality rules.
The OECD's 2025 Common Reporting Standard concerns automatic exchange of information on financial accounts among participating jurisdictions. Its official overview records the adoption of the original standard in 2014 and an annual exchange framework. Such reporting can improve the capacity of competent authorities to identify cross-border account relationships. It does not itself create a publicly accessible census of every security and beneficial owner, or establish completeness of international portfolio data.
A reporting requirement directed to financial institutions and a survey of national portfolio assets have different units. Account-level information may provide values or income associated with reportable persons, whereas a position dataset links particular financial instruments and institutional sectors with national accounts. Reconciling them requires common definitions of values, dates, holder status, and the scope of instruments. A reported account balance may include nonsecurity assets. Adding it to the position already observed by a portfolio survey without linkage risks duplicate counting.
From an economic policy perspective, transparency can reduce harmful asymmetry between private institutions that observe financial claims and public authorities with legitimate information requirements. It can also impose administrative costs and privacy risks. The design question is which information is necessary, who has authority to obtain it, how it is securely used, and how aggregate findings can be disclosed without exposing individuals. A system designed for analytical precision therefore includes confidentiality controls rather than attempting unrestricted disclosure.
Reforms may change observed data by improving the coverage of existing assets rather than changing the underlying allocation. Consider a survey that introduces better reporting in year t and discovers previously unobserved positions of 20. Even if those securities were purchased years earlier, current reported stocks rise. Treating the entire recognition as newly acquired foreign securities misidentifies the timing of transactions. The same issue applies to a newly reported account under tax transparency reforms. Evaluations should therefore distinguish new economic activity from the reporting of previously existing positions.
Policy effectiveness requires appropriate outcomes. A fall in the measured stock in a particular center might reflect divestment, migration to another jurisdiction, market repricing, or changed survey practice. An increase in reported assets may indicate improved compliance rather than greater tax evasion. Credible evaluation of reforms demands a design that identifies relevant counterfactual trends and reporting changes. This dossier draws implications for measurement, not jurisdiction-specific legal advice or a causal estimate of transparency policy.
Research design, identification, and reproducibility
Empirical research should define the estimand before assembling a dataset. Possible estimands include the market value of resident investors' portfolio claims on nonresident issuers, the proportion of these claims issued in a pre-specified set of financial centers, the amount reallocated to a particular parent economy, the distribution of fund underlying assets, or an estimate of unidentified household securities ownership. Each requires a different observational and inferential framework.
For a residence-based concentration ratio, specify the universe P of directly observed positions for a particular reporting economy and period, and a fixed set C of issuer financial-center residences. Divide the market value of positions with issuer residence in C by the total market value of P, if P has a positive total and consistent long-position treatment. This is an issuer-location share. It measures neither the investor's use of foreign custody nor the ultimate parent concentration. For data including short positions, the analyst should separately specify gross long, gross short and net amounts, because net denominators can become small or negative.
For a parent exposure study, retain original residence assignments and construct a crosswalk from instruments to parent entities and countries. Version the crosswalk at each observation date, because mergers, inversions and reorganizations may alter group identity. Record unmatched securities and the value of ambiguous assignments. Where several plausible allocations exist, report sensitivity to the alternatives. Validation should compare group assignments with independently available issuer records for a sample and describe why exceptions occur.
For estimating changes through time, establish a common-price, common-currency and common-scope decomposition where feasible. Revisions of published data should be retained with source vintage. Changes in the number of reporting economies or institutions alter the coverage of the matrix. A balanced panel can help identify changes within a constant reporter universe but may underrepresent newly important jurisdictions; a broader panel can improve contemporary coverage while creating historical breaks. Presenting both can reveal the sensitivity of results to sample design.
Constructing a global ownership matrix is difficult because security identifiers can be repeated through reporting chains, and access to end-investor and issuer data is not uniform. A reproducible protocol needs rules for unique records, reporting duplicates, valuation, handling of multiple asset and custody levels, and the allocation of unknowns. An independently observed aggregate total is a useful constraint, but matching it does not validate the distribution across investor economies.
Causal estimates require a separate identification strategy. For example, a reform affecting one class of offshore securities could be studied using comparable unaffected instruments, but the groups must have defensible pre-reform trends and no unaddressed contemporaneous changes in reporting, investor demand, or market access. Spillovers may contaminate comparison groups if investors substitute between vehicles. A simple before-and-after comparison of positions would combine these mechanisms. A documented event-study or other suitable design can improve interpretation where the data support it.
The present research is a source-based methodological synthesis. It does not fit a causal model or estimate a new global matrix. Its accounting demonstrations are transparent and can be reproduced manually. The literature's estimates are not independently replicated here. Access to several journal articles and methodological chapters was limited to indexed abstracts or publication excerpts, restricting source-specific assertions accordingly. This limitation is substantive for detailed replication or methodological comparison; it is less consequential for the demonstrated accounting identities. A future empirical extension would need the underlying vintages, instruments, parent mappings, and permissions to use confidential data.
Analytical synthesis and implications
The evidence supports several interconnected economic conclusions. Offshore securities holdings are multidimensional because financial claims, ownership, custody, funding and production have separable institutional geographies. The primary economic problem is selecting a measurement object and reconciling it with the relevant data. The researcher's choice between separate legal-unit balance sheets and consolidated exposures must follow the question being asked.
The first implication concerns international financial integration. Residence accounts identify relationships between institutional units, while consolidated parent and revenue mappings describe alternative connections between investors and economic activity. Both datasets can be internally consistent and yet generate different country rankings. Changes between them measure reclassification under stated assumptions, not necessarily new transactions.
The second implication concerns the economic importance of specialized jurisdictions. Large external financial positions may arise from intermediation of capital that is generated, managed or deployed elsewhere. Assessing the benefits and vulnerabilities to a financial center's economy requires separating fees, value added and domestic fiscal consequences from client asset values and the balance sheets of locally domiciled vehicles. A small economy can provide large-scale international financial infrastructure while retaining relatively modest domestic net wealth.
The third implication concerns global distribution. Missing household assets are a possible explanation for discrepancies in international portfolio statistics, but determining their size and ownership requires independent evidence about data coverage and valuation. An estimate of offshore household wealth is a different statistic from the value of securities issued or managed via financial centers. Distinguishing these measures prevents both understatement of omitted assets and overstatement through duplication.
The fourth implication concerns systemic risk. Consolidating securities positions by group may reveal common borrower exposure, while underlying portfolio and currency data reveal additional vulnerability channels. Legal obligations, contingent support, collateral, redemption terms, and financial derivatives determine how a disruption propagates. No one geographical column can establish all these dimensions.
The fifth implication concerns public administration. Better data integration demands agreement on identifiers, instrument classes, populations, reporting dates and valuation, alongside lawful access and protection of confidential information. It also demands explicit preservation of source vintages and assumptions. An apparently complete matrix created by imputing every unknown without documenting uncertainty can be less useful than a matrix that accurately marks what is unobserved.
A useful operational conclusion is to publish parallel statistical views where resources and confidentiality permit: the residence-based institutional position, issuer group exposure, instrument and currency distribution, and the coverage and uncertainty of deeper look-through. These views should share a documented reconciliation and retain estimates as distinguishable from observed quantities. This is a practical recommendation derived from the reviewed literature and accounting analysis; it is not advanced as a new formal economic theory.
The central challenge of offshore economics is not that financial centers create multiple equally valid measures of a single undifferentiated quantity. It is that their intermediation relates several real economic quantities. A rigorous account preserves the difference between those quantities and makes explicit the transformations required to move between them.
Conclusion
Offshore securities holdings are a significant object of international economic research because they reveal the divergence between the institutional location of financial contracts and the geography of the capital, activities and risks connected through those contracts. Their measurement requires instrument-specific balance-sheet analysis, precise geographical attribution, and explicit treatment of missing information.
The research establishes why custody location should be separated from economic ownership; why issuer residence should be retained even when parent nationality is informative; why fund shares and underlying assets require explicit consolidation; and why positions must be distinguished from transactions, valuation changes, and reclassification. Historical empirical investigations of U.S. portfolios and global corporate issuance demonstrate that alternative allocations can materially alter the observed geography of investment. Household offshore-wealth estimates answer a related but separately defined question.
The policy consequence is the need for transparent, purpose-specific measurement. International statistics, prudential information and tax disclosures can be complementary, provided their distinct populations, objectives and legal frameworks are respected. Analysts should publish source coverage, uncertainty, and the method of each reallocation rather than interpret a financial-center label as proof of the nature or intention of individual ownership.
The original contribution of this dossier is the synthesis of these statistical, institutional and financial perspectives, including conservation of portfolio value under defined allocation weights, illustrative stock reconciliation, separation of gross and net fund exposures, and an explicit bounded approach to unresolved investor geography. These are applications of established accounting and analytical reasoning, not claims of theoretical priority or new empirical findings.
Reliable offshore securities analysis ultimately depends on a simple discipline: identify the claim, identify the holder and issuer at the correct institutional level, identify the transformation used to infer alternative economic exposure, and preserve every material distinction when interpreting the resulting aggregates. The offshore system becomes intelligible through these linked, auditable measurements.