RESEARCH DOSSIER

Conduit and Sink Offshore Financial Centers

Abstract. This monograph examines the economic organization of conduit and sink offshore financial centers through the published corporate ownership network taxonomy, foreign direct investment statistics, securities reallocation, banking geography, institutional specialization and fiscal evidence. It reconstructs the distinction between GDP-normalized net sink centrality and conditional intermediate conduit centrality, examines their interpretation and sensitivity through hypothetical calculations, and distinguishes ownership-linked weights from cash flows, investment positions, profit and productive activity. Comparative research on offshore FDI and financial intermediation establishes the importance of institutional complementarity and measurement perspective. The synthesis identifies implications for geographical monitoring, investment analysis, tax policy and financial stability. It preserves the historical scope of reported classifications and treats the independent replication of proprietary network data and causal estimation as separate empirical tasks.

Keywords: Offshore financial centers; conduit OFCs; sink OFCs; corporate ownership; financial geography; network centrality; foreign direct investment; special purpose entities; pass-through funds; offshore economics; international financial statistics

Introduction

Conduit and sink offshore financial centers are specialized positions within the geographical organization of international corporate ownership. A conduit occupies an intermediate position through which ownership-linked value is attributed between economies; a sink is an economy in which the net weight of incoming ownership-linked value is disproportionate to the size of its domestic economy under a specified empirical classification. Both roles can coexist in a jurisdiction. The subject of this monograph is the economic meaning of those positions, their measurement, their institutional foundations, and the consequences of interpreting global investment through them.

The central problem is that the location of a legal claim and the location of the economic activity associated with it are separable. A multinational enterprise can have operations, funding, management, shareholders, intellectual property and creditors in different economies. In a consolidated report some intragroup relations disappear; in national external accounts those same legal relations may constitute genuine cross-border financial positions. Each view answers a different question. The economic geography of the multinational enterprise therefore cannot be inferred from a single map of financial assets.

This study develops an analytical synthesis of published evidence. Its first object is the corporate-ownership taxonomy introduced by Garcia-Bernardo et al. (2017) in Uncovering Offshore Financial Centers. It then assesses how the taxonomy relates to direct investment, securities issuance, bank reporting and profit allocation. The scientific contribution is an interpretation of the correspondence and limits among these measures, with independently constructed arithmetic examples and implications for monitoring. There is no claim to discovery of the conduit/sink distinction, original estimates of global rankings, or a newly established general theory.

The owner Science is Offshore Economics (EBS-081). General offshore financial networks concern multilayer system topology; the present dossier concentrates on intermediate and terminal roles in ownership paths and their economic interpretation. Classification of financial centers in general, sectoral banking and fund economics, doctrinal tax and corporate law, insurance regulation, and the psychology of offshore activity remain separate research objects. The economic analysis is global rather than a legal or investment recommendation concerning any particular jurisdiction.

The economic object and the meaning of offshore

Economic activity becomes offshore in a relational sense: a claim or institutional service is organized in a jurisdiction other than that of a relevant owner, user, operating enterprise or underlying economic exposure. These possible reference points must be stated, because they do not necessarily coincide. A domestic bank booking a foreign-currency transaction and a foreign-owned holding company registered locally each participate in international finance, but the basis of their offshore character differs.

Zoromé (2007) proposed an operational definition of OFCs centered on services supplied to nonresidents at a scale unusual relative to the local economy. This serves as a functional starting point. An economy's cross-border specialization may support substantial local expertise, infrastructure and employment even if the nominal financial claims passing through it are much larger than its domestic production. A size-normalized financial position is consequently a measure of specialization, not a direct measure of domestic dependence, social welfare or misconduct.

Three dimensions merit separate examination. The first is scale: the absolute amount of internationally oriented financial intermediation. The second is intensity: the relative importance of such activity against an economic denominator. The third is organization: whether a center predominantly records endpoints, intermediate claims, or a combination of functions within a specified network. Scale helps assess the potential systemic importance of concentration. Intensity helps detect small economies whose external positions exceed what their size alone would suggest. Organization identifies the structural relationships that make one jurisdiction complementary to another.

Consider two hypothetical economies with equal external assets. The smaller economy may have much higher offshore intensity. If the larger economy also intermediates many paths linking third countries, it may be an important conduit despite its lower intensity. Conversely, a small economy can have substantial foreign asset holdings while connecting relatively few third-country paths. Replacing any of these three dimensions with another loses information.

A jurisdiction is not a unitary financial institution. An economy can include locally owned banks, foreign branches, collective investment funds, holding companies, operating subsidiaries and households. Each may have different counterparties and reasons for its international connections. Classification at country level summarizes the chosen observations; it is neither a finding concerning every resident nor a judgment of every legal arrangement. A scientifically defensible statement attaches a role to an identified data layer, period, method and economic denominator.

The term sink refers here to a direction of attribution in an ownership network. It does not imply that money physically remains in a jurisdiction, that financial claims are illiquid, that the ultimate beneficiaries reside there, or that all activity ceases at that location. Similarly, a conduit need not have a zero net international investment position. A node can intermediate claims while independently accumulating positions and providing local services. These distinctions are indispensable to the empirical interpretation that follows.

Corporate ownership as a weighted directed network

Firm-level ownership data make visible institutional links that disappear when international statistics are reduced to bilateral totals. The graph represents legal entities as nodes and ownership ties as directed edges. An edge can carry an ownership share, an instrument or a modeled control weight depending on the data and research question. Vitali, Glattfelder, and Battiston (2011) use a weighted ownership graph to investigate global corporate control. Their work provides background for distinguishing economic ownership from the exercise of control.

Suppose corporation X is an operating enterprise, entity Y holds shares in X, and entity Z owns shares in Y. In the orientation used in the conduit/sink literature, the path runs from X to Y to Z, toward owners. A graph that instead directs edges from investor to investee reverses the representation. Before comparing results across studies, an investigator must check the orientation. A sign reversal in a measure of net inflow can otherwise create a superficially plausible but fundamentally opposite conclusion.

If Y owns 60 percent of X and Z owns 50 percent of Y, the multiplicative interest of Z along that single chain is 30 percent. This path calculation measures proportional exposure through that sequence of ownership, and does not establish Z's voting control. Several parallel chains may lead to the same entity; shares in multiple chains can overlap or be held through reciprocal arrangements. Effective control also depends on voting rights, dispersion among other shareholders, contractual arrangements and corporate governance. Consolidating these facts into one edge weight requires a specified method.

Country-level inference adds another operation. A chain may contain several entities in the same economy. Compressing contiguous domestic segments can clarify cross-border routing without deleting the corresponding enterprises from the underlying dataset. But assigning an economy using legal domicile rather than headquarters or effective management changes the geography. For an analysis of offshore registration, this is often appropriate; for an analysis of operational output, an alternative classification may be more meaningful.

Path membership and actual transfers are independent observations. A company can retain earnings, issue securities, reorganize equity ownership, or change debt without a contemporaneous payment along every ownership edge. The graph carries information about structure and potential economic attribution. Bank payment records or accounting distribution data would be required to trace an actual transfer. Even a complete graph of registered ownership would not automatically reveal the date or amount of every dividend, interest payment or cash movement.

There is a direct implication for terminology. In this monograph, ownership-linked value denotes an analytical weight associated with a chain. It is distinct from equity market value, enterprise value, FDI position, profit, dividend or liquid wealth. The word route describes an observed or modeled sequence of institutional relationships unless a transactional dataset separately establishes a flow. These are not stylistic qualifications: they determine the units in a quantitative claim.

The published conduit and sink methodology

Garcia-Bernardo et al. (2017) analyze large-scale ownership data encompassing more than 98 million entities and 71 million ownership relations. Their firm-level chain construction uses a depth-first traversal, a multiplicative ownership cutoff of 0.001, cycle avoidance, and revenue times multiplicative ownership as the path weight. They aggregate consecutive same-country entities and construct shorter jurisdictional motifs for their centrality measures.

Let G2 be the collection of aggregated two-economy motifs and G3 the corresponding three-economy motifs. Write v(g) for the weight of motif g, D(c) for GDP of economy c, and D for the sum of GDP across the comparison set. Define I(c) as the sum of G2 weights for which c is the second economy and O(c) as the sum for which c is first. Let T2 be the sum of all G2 weights. The sink centrality reported in the study can be expressed as

S(c) = ((I(c) − O(c)) / T2) × (D / D(c)).

This is a signed dimensionless statistic. The net incoming share of the network is divided by the economy's GDP share. Its sign identifies the direction of net attribution. The study classifies sink OFCs using a threshold of 10. Such a threshold is an empirical convention of that investigation, not a universal legal criterion. An economy whose GDP is small can attain a large S with a modest absolute net incoming weight.

The conduit calculation conditions on the previously identified sinks. Define W_in(c) as the weight of three-country motifs starting in a sink and containing c as the middle economy. Define W_out(c) as the counterpart weight for motifs ending in a sink. With T3 equal to the aggregate weight of the three-economy motifs, the two ratios are

C_in(c) = (W_in(c) / T3) × (D / D(c)),

C_out(c) = (W_out(c) / T3) × (D / D(c)).

The published criterion uses both ratios above one. Because the motifs are oriented toward owners, the words inward and outward identify the source paper's conventions, not the direction of cash or FDI financing. The two axes preserve information that would disappear if they were prematurely averaged.

The use of two- and three-country motifs gives the study a specific structural interpretation. A classification of intermediate locations is not conventional shortest-path betweenness centrality. It depends on observed ownership chains, their weights, and their relationships to a selected set of sinks. A country can have considerable generic network centrality without meeting this conditional criterion.

Two mathematical consequences follow from the definition. First, the sum over countries of GDP-share-weighted signed sink centralities is zero for a complete, internally consistent two-country motif collection: total incoming motif weight equals total outgoing motif weight. This is an accounting identity of the measure, not a claim about the world's balance of payments. Second, the same global denominator and the same domestic GDP ratio are used for each conduit axis, so their ratio, when the outward value is positive, is W_in divided by W_out. This directional asymmetry can be discussed separately from the size normalization.

The calculation is sensitive to coverage, weighting, domestic compression, thresholding and choice of comparison set. It is important that interpretations retain their dependence on the published construction. The present monograph reconstructs the algebra of the published measure; it does not recreate the original database, re-estimate the scores or verify the underlying firm records independently.

Worked examples and sensitivity

A transparent numerical example makes the distinction between concentration and intermediate routing explicit. All values below are fictional and in arbitrary consistent monetary units. They are not estimates for any real economy.

Suppose total two-country motif weight is 1,000 and the total GDP of all economies is 2,000. Economy A has GDP of 20, incoming motif weight of 135 and outgoing motif weight of 15. Its net attribution is 120. Dividing 120 by 1,000 gives 0.12; multiplying by 2,000 divided by 20 gives a sink centrality of 12. If A's GDP were 40 while all other inputs and the world denominator were held fixed for a partial sensitivity calculation, the score would be 6. The observed path structure would be unchanged, but classification under the published threshold would change. In a complete comparative recalculation the world GDP denominator should also be updated where appropriate.

Economy B has a GDP of 100, incoming weight of 105 and outgoing weight of 100. Its sink centrality is 0.1. B is approximately balanced on this measure even though it has substantial gross links. Suppose the total weight of three-economy motifs is 500, and B is the intermediate economy in sink-origin motifs totaling 35 and sink-ending motifs totaling 45. Its inward conduit centrality is (35/500) × 20 = 1.4, and its outward conduit centrality is (45/500) × 20 = 1.8. It would meet the published conduit criterion for these hypothetical data.

A third economy, C, receives 70 and sends 10, with GDP of 10. Its signed score is (60/1,000) × 200 = 12. A and C therefore have identical normalized centrality but different absolute net weights. A researcher interested in systemic concentration should report both. An analyst who ranks them only by the score would conceal the absolute difference, while one who ranks solely by the nominal amount would conceal C's extremely high comparative intensity.

The examples demonstrate that a conduit may have limited net accumulation and a sink may be small in absolute financial volume. They also show why a classification boundary is not a qualitative discontinuity in economic behavior. A score of 9.9 and one of 10.1 may describe nearly identical underlying configurations. The numerical threshold helps organize an empirical study; decisions based on it require sensitivity tests and the associated gross quantities.

A weighted example clarifies the influence of the chosen valuation variable. Assume initial operating revenue of 200, a first ownership link of 80 percent and a second of 50 percent. The path weight is 200 × 0.8 × 0.5 = 80. If revenue rises to 300 without a change in corporate ownership, the weight rises to 120. The graph is unchanged, yet its weighted centrality may shift. Thus the distinction between organizational change and change in the scale of operating activity should be retained in longitudinal analysis.

Conversely, when revenues are unchanged and a holding entity relocates from one jurisdiction to another, the attributed pathway can change without a change in productive output. The graph may correctly register a structural reorganization, but the implication for investment in factories and workers would need separate evidence. In both cases the change is economically informative; the reason for the change differs.

A useful robustness exercise would calculate exact results for a fixed coherent panel, alternative positive GDP denominators, a range of predefined sink cutoffs, and alternative valuation measures when available. Each alternative would be a sensitivity scenario whose results are reported rather than silently substituted into the baseline. The present illustrative calculations verify units and signs only. They do not constitute a statistical sample or a new empirical estimate.

Empirical findings and geographical specialization

The original 2017 corporate ownership study reports 24 sink jurisdictions and five major distinct conduit economies: the Netherlands, the United Kingdom, Switzerland, Singapore and Ireland. Belgium, Panama and Guernsey appear as smaller distinct conduits; some detected sinks also have important conduit roles. This is a historical study result, not a current ranking.

The empirical importance of these results lies in the geographical composition of the set. Industrialized economies and large service centers connect a substantial part of corporate offshore activity. An interpretation organized exclusively around small island jurisdictions would miss intermediate relationships elsewhere. Conversely, treating the large centers merely as generic locations of financial services would miss the specific importance of their offshore-linked ownership connections.

Institutional specialization can coexist with diversification. A large center may provide commercial banking, legal services, corporate headquarters functions, investment management, trading infrastructure and ownership intermediation. Some of these activities are internationally oriented and some support domestic production. A large observed conduit score does not establish the proportion of domestic output arising from a particular category. For that question, value-added, employment and service-export data are needed.

The geographical counterpart structure is a different attribute from total weight. Two conduits can record the same aggregate weight while serving markedly different counterpart economies. One may provide a broadly distributed service to many groups; another may specialize in a narrower regional or sectoral relationship. Neither configuration alone determines whether entry barriers or substitution costs are high. The relevant evidence includes contracts, business organization, regulation, professional markets and the availability of alternative centers.

Haberly and Wójcik (2015) examine regional blocks and imperial legacies in offshore FDI. Their geographical perspective helps explain why cross-border networks exhibit persistent regional patterns rather than a uniform distribution governed by size or tax rates alone. The implication for conduit analysis is that differences in history, language, institutional familiarity and standardization may be confounded with differences in fiscal regimes. A causal claim identifying one of those influences requires a credible strategy to separate it from the others.

An appropriate empirical comparison therefore involves both scale and composition. Absolute motif weight addresses how much observed corporate organization depends on a location. Normalized weight addresses disproportion relative to economic size. Counterpart distribution addresses the diversity and concentration of potential relationships. Sectoral composition addresses the kinds of institutions that occupy the relevant positions. None is a substitute for the remaining variables.

The geographical results motivate research into historical persistence, but historical persistence must itself be demonstrated with comparable observations through time. A change between two lists assembled with different data vendors, entity definitions or inclusion rules does not identify migration. A panel constructed using constant boundaries and comparable entity coverage would be better suited to that question.

Specialization, institutional complements, and local economic activity

The organizational economics of financial centers begins with specialization and coordination costs. A jurisdiction becomes attractive to international actors when it offers a sufficiently useful bundle of institutional services. These may include predictable adjudication, standardized contracts, professional expertise, reliable reporting, currency access, fund administration, governance frameworks and the ability to interact with counterparties elsewhere. Such characteristics can support genuine intermediation even where financial claims associated with the local entities exceed local production.

A holding entity may simplify consolidated organization or separate the responsibilities of groups of investors. A financing subsidiary may centralize liability management. A regulated fund can pool investors from multiple markets under a single governance arrangement. The economic benefit of these structures depends on what functions they perform, the costs they create, and how risks and responsibilities are allocated. The existence of a chain is evidence of organization. Determining its economic contribution requires evidence of actual services and outcomes.

Fichtner's (2016) Cayman Islands case study connects offshore financial specialization with international hedge-fund activity and relationships with major external financial economies. The case illustrates why a financial center may be closely related to capital owners, managers and market infrastructure in other jurisdictions. The institutional role recorded in a fund or holding-vehicle jurisdiction need not identify the residence of the investors or the location of security selection.

An analytical distinction between fixed and variable costs is useful. Infrastructure, institutional reputation and trained specialized labor often involve fixed costs. The marginal cost of administering an additional external entity may be relatively low when processes are standardized. Scale can therefore sustain a specialization that would be uneconomical in a dispersed arrangement. The same mechanism may create concentration and dependencies. The distribution of economic gains between providers, investors, workers and governments depends on market structure and incidence rather than on the number of entities alone.

Jurisdictions compete over more than nominal taxes. They also compete over predictability, administrative delay, access to service networks, dispute resolution and compliance burdens. A reduction in one class of costs can be offset by increases in others. Firms and financial institutions may maintain existing structures even when a new alternative appears cheaper if switching introduces uncertainty, contractual changes or operational risk. This creates possible path dependence without implying that every existing arrangement is optimal.

The conduit and sink roles can reinforce each other because intermediary and endpoint services may be complements. If an endpoint provides specialized corporate or fund administration, its accessibility may increase with the existence of reliable intermediaries. If a conduit provides connectivity, its usefulness may depend on the range of underlying assets, counterpart institutions and markets it can connect. The relationship is conditional on the services in question and cannot be assumed for all offshore centers.

Local economic value is best measured by locally produced services, labor income, government revenue and other appropriate national accounting concepts. Counting the nominal assets administered by a jurisdiction as its output would inflate the apparent scale of local production. At the same time, dismissing the entire service sector because the assets are externally owned would understate the real economic functions it performs. Economic analysis must preserve both facts.

Foreign direct investment and the problem of pass-through

Foreign direct investment is a statistical category of cross-border financial relationships. Its bilateral geography ordinarily reflects specific resident institutional units. When the ownership and financing relationships within a corporate group cross multiple boundaries, a recorded inward position can be accompanied by a related outward position. A large gross position in an intermediary economy can therefore coexist with modest domestic operational investment associated with those particular entities.

Damgaard, Elkjaer, and Johannesen (2019), in What Is Real and What Is Not, estimate that phantom FDI may represent almost 40 percent of global FDI in their studied network. They report that reallocating genuine investment to ultimate investor economies and removing phantom positions improves the explanatory performance of standard gravity variables. The result concerns the study's FDI estimand, not the revenue-weighted ownership centrality measure and not a contemporary estimate as of this monograph's date.

A simplified balance-sheet example illustrates why. Suppose a parent in P contributes 100 in equity to a holding entity in C, which contributes all 100 to an operating subsidiary in O. Under a simple residence-based presentation, P has a cross-border asset in C, while C has a corresponding liability and an asset in O. O has a liability to C. Each is a distinct institutional claim. Adding the two successive investment links gives 200 in gross positions associated with this simplified chain, although the final operating subsidiary receives 100. This is a constructed equity example assuming equal book values, no additional funding, no valuation change and no separate assets.

An ultimate-parent view identifies the originating group relationship between P and O. It is invaluable for particular questions of corporate connection but does not extinguish the legal claims located in C. The residence view and the ultimate-parent view answer complementary questions: the former describes the distribution of cross-border institutional claims; the latter identifies relationships obscured by intermediate entities.

This example also explains why the term double counting needs careful use. For the measurement of gross external claims, the two links are legitimate observations and counting both is not an error. For the measurement of unique ultimate investment reaching operations, counting 200 would misrepresent the chosen economic object. The problem is one of estimand definition.

A negative net position in the intermediary economy cannot by itself be interpreted as a loss of investment. In practice, exchange rates, transfers of equity, changes in market valuation, retained earnings, financing structure and other factors can affect the relationship between inward and outward positions. A reconciliation requires dates, instruments and sufficiently detailed accounts.

The fifth OECD benchmark definition (OECD 2025) expands guidance on purpose of investment, ultimate investing and host economies, and identification of pass-through funds. It is a statistical improvement precisely because users need several views rather than a single supposedly definitive geographical allocation. Historical series should distinguish the applicable compilation standard and its implementation date.

The analytical distinction between financial routing and productive investment also matters for development policy. A country may report a substantial rise in inward investment during a corporate reorganization that creates little new local productive capacity. Another country may receive a productive project financed through a foreign intermediary whose identity dominates the bilateral investment record. Evaluation of employment, output, productivity or technology transfer should consequently draw on project, enterprise and activity data as appropriate rather than rely solely on headline bilateral FDI.

Special purpose entities and the limits of institutional proxies

A special purpose entity is a statistical and institutional category with defined characteristics. The OECD treatment of special cases (2025, chapter 6) identifies SPEs using factors including institutional-unit status, foreign control and very limited local presence. It also distinguishes pass-through funding from SPE status, because operating entities can intermediate resources.

This distinction is central to the subject. A jurisdiction is a location; an SPE is an entity; a conduit is a role in a network; pass-through is a property of funds or positions under an accounting treatment. These terms operate at different levels of observation. A jurisdiction with many SPEs need not be highly central in the corporate ownership network relative to its size. A conduit can contain ordinary operating firms, and an entity formally categorized as an SPE can be an endpoint of a particular observed sequence.

Using entity flags as the sole operational definition of a conduit creates false negatives when intermediary functions are performed by active businesses. It creates false positives when entities categorized as SPEs perform functions outside the particular path under examination. The converse error is to infer SPE status merely because a firm occurs in a path through a jurisdiction identified as a conduit. Both classification errors can affect aggregate estimates of local activity.

The IMF's SPE database launch in 2022 provided cross-border data initially covering 25 economies, with annual observations for 2020 and additional periods for some participants. That historical coverage statement should not be interpreted as a current count. The usefulness of a harmonized SPE dataset depends on classification consistency, completeness, treatment of domestically related entities and the date of reporting.

A comparison with ownership graphs would require matching the criteria and reference periods used by each source. A corporate database may identify an address or registration location but omit information sufficient to apply the statistical definition of an SPE. Conversely, official aggregate statistics may separate SPEs but suppress detailed counterparty cells for confidentiality. Neither limitation is necessarily an error in the underlying dataset, because their purposes differ.

For policy analysis, an increase in entity substance may involve real additional services or merely changes in measurable characteristics. Reliable assessment requires evidence about functions performed, personnel, assets, risks and activity rather than a formal label alone. The legal adequacy of substance under a specific regime is a distinct legal question. Economic analysis here concerns the interpretation of data and observable activities, not instructions for meeting or circumventing specific requirements.

Securities issuance, financial risk, and ultimate parents

A separate geography appears when companies finance themselves through debt or equity securities issued by subsidiaries. The immediate issuing entity can be located in a jurisdiction different from that in which the parent group is headquartered. In a security-level dataset this issuer residence is a meaningful identifier even when the underlying economic exposure is better understood at a group level.

Coppola, Maggiori, Neiman, and Schreger (2021) reallocate traded securities to ultimate parents. Their research shows how this adjustment can materially change bilateral portfolio investment estimates. Their historical U.S.–China illustration identifies an understatement of nearly $600 billion in U.S. positions in Chinese firms under the conventional geographical view. The magnitude is specific to that study and must not be combined directly with the phantom-FDI percentage.

Consider an investor in economy A holding 100 of debt issued by an entity in economy C. The issuing company belongs to a group headquartered in H and finances activities elsewhere. An issuer-residence statistic records A's holding against C; a nationality-based analytical reconstruction may associate the position with H. The full contractual credit risk may additionally depend on guarantees, collateral, seniority, cash flows, insolvency arrangements and the actual structure of the group. Ultimate parent and ultimate risk are therefore distinct analytical concepts.

An intermediary issuer may be essential for transaction execution while having limited operational output. Identifying its parent improves understanding of common economic exposure, but the residence of the claimant and debtor remains important for jurisdiction-specific supervision and legal enforcement. Risk can be transmitted through both the legal chain and the consolidated balance sheet. The two representations are complementary.

By analogy, a country appearing as an ownership sink in the corporate graph may be a significant issuer domicile yet have a very different position in a nationality-based securities map. It may also host investors buying instruments from other countries. These activities do not cancel in a single economic sense. The records describe different assets and liabilities whose distribution matters for financial analysis.

A research design comparing securities and ownership would ideally match legal-entity identifiers and parent mappings by instrument and time. Corporate reorganizations can change parent attribution; changes in security holdings can occur without changes in issuer identity. A static match used for all historical periods can introduce look-ahead error. The match should be versioned, and unresolved entities should be reported rather than allocated arbitrarily.

Banking geography and the identification of financial centers

The BIS locational banking statistics organize banking activity by the location of reporting offices. The consolidated banking statistics approach banking from the nationality perspective of parent groups. These views are valuable for different purposes. A locational record helps describe where a claim is booked; a consolidated view helps describe exposures associated with international banking groups.

Imagine a banking group headquartered in H with an affiliate booking a loan in C to a borrower resident in B. Residence-based data may describe the exposure as originating from C. Consolidated data can associate it with the group headquartered in H. If the borrower carries out operations in additional jurisdictions or receives a guarantee from another entity, even these two locations need not exhaust the question of ultimate economic risk.

The interpretation of a conduit in a corporate ownership network must not be transferred automatically to banking. A wholesale banking office can intermediate funding without being a central holding-company location. Conversely, a jurisdiction with a large stock of legal entities may not host a banking system with equally extensive credit positions. An economic account of offshore specialization should retain layer-specific roles.

Pogliani and Wooldridge (2022) focus on cross-border financial centers as a distinct object for monitoring international financial relationships. Their work supports the need to distinguish specialization in nonresident business from classifications based exclusively on tax or regulatory characteristics. An internationally oriented banking center may also serve a domestic economy and provide market infrastructure.

A combined geographical monitoring system would place residence, banking nationality and corporate ownership observations alongside each other with clearly identified denominators. A jurisdiction's contribution to external intermediation could be high under one indicator and low under another. This divergence can indicate institutional specialization rather than an anomaly needing statistical correction.

For financial-stability work, the amount of observed claims is only an initial input. Funding maturity, liquidity, leverage, guarantees, netting and hedging conditions shape vulnerability. Market disruptions may propagate across contracts that do not follow the owner-directed corporate graph. A center's corporate conduit score is thus insufficient to quantify systemic risk. It can nevertheless help identify where additional balance-sheet and contractual investigation would be warranted.

Profit allocation and fiscal interpretation

Offshore ownership arrangements can affect the geography of taxable income, but the amount of corporate ownership-linked value in a jurisdiction is not a measure of profit shifted to that jurisdiction. Profit is a flow of income over a period; an equity position is a stock; the path weights analyzed above are calculated from company operating revenues and ownership shares. The units may all be expressed in currency, yet the quantities have different economic meanings.

Tørsløv, Wier, and Zucman (2023) estimate in The Missing Profits of Nations that 36 percent of multinational profits are shifted to tax havens under their empirical methodology. The published estimate concerns the allocation of multinational profit, not a share of global assets, FDI stocks or all corporate revenue. It must be read within the assumptions and coverage of that study. No estimate of fiscal loss follows from applying this percentage to the conduit scores.

Suppose an operating enterprise records 200 in revenue, 120 in operating costs and 20 in other deductible costs before tax for a hypothetical period. Its illustrative profit is 60. A holding company's ownership interest may produce a revenue-based network attribution of 100 if its stake is 50 percent. That figure is neither the enterprise's profit nor the amount of profit shifted. To estimate a tax consequence, an analyst would need the applicable profit allocation, tax base, counterfactual and treatment of the relevant income and expenses.

The fiscal effects of a corporate organization also differ by country and stakeholder. A source jurisdiction may experience changes in its taxable income. A financial center may receive taxes, registration fees, professional-service income and other revenues. The parent jurisdiction may tax certain income on a different basis. Investors and employees may bear part of the resulting incidence through changes in returns and remuneration. Estimation of net welfare effects must consider responses as well as these accounting entries.

The empirical link between network organization and profit allocation is a question for research. A correlation between a conduit's score and low effective taxation could support further investigation, but it cannot by itself identify a causal effect. Some activities would have been organized internationally even in the absence of a tax difference; other arrangements may respond strongly to fiscal incentives. A credible counterfactual separates these mechanisms.

For an individual enterprise, evidence of illegality requires applicable law and case-specific facts. Inferences from network measures operate at an aggregate statistical level. They are appropriate for identifying concentrations and forming hypotheses; they do not establish evasion, money laundering, beneficial ownership or liability. The monograph treats fiscal and public-policy risks as economic objects and preserves this distinction throughout.

Volatility, historical series, and the interpretation of headline investment

International investment aggregates can vary substantially when financial routing changes. A year with large restructuring, changes in intragroup financing or liquidation of holding positions may produce a headline investment pattern quite different from the underlying change in productive activity. This is particularly important when gross and net flows are combined in regional comparisons.

UNCTAD's January 2025 Global Investment Trends Monitor No. 48 provides a striking example from the preliminary 2024 vintage: it estimated an 11 percent increase in headline global FDI and an 8 percent decline after excluding flows through European conduit economies. These figures describe the published preliminary release, not a final revised estimate or a contemporaneous ranking of conduit centers. Their disagreement demonstrates the sensitivity of an annual interpretation to the treatment of financial routing.

A rigorous time-series analysis should record at least the year of the observation, data vintage, measured instrument, sign and accounting principle, and any exclusions. This information is particularly useful when negative flows represent withdrawals, repayments or restructuring. Replacing negative values with zero might simplify a chart but would distort a net aggregate.

The relationship between headline investment and conduit-adjusted investment should be explained in terms of the publication's exact exclusion convention. Removing selected economies is a practical descriptive adjustment and can help identify a different broad trend. It is not equivalent to tracing every ultimate investor and final operating recipient or individually measuring pass-through funds. The adjustment may exclude genuine domestic investment in the affected economies, and some intermediary positions elsewhere may remain.

A time series of ownership scores also requires stable firm coverage. A database expansion can increase the number of visible links and weighted paths without any actual change in global corporate organization. Conversely, a reporting gap can reduce measured centrality. The observation of missingness should distinguish missing entities, missing ownership percentages, missing operating revenues and missing sector codes. Different forms of missingness have different implications.

Comparability through time is consequently more valuable than a superficially comprehensive but inconsistent set of snapshots. For policy evaluation an analyst should reconstruct a comparable pre-reform and post-reform sample and test whether changes are robust to shifts in coverage. Relevant contextual changes include market cycles, exchange rates, reorganizations, corporate entry and exit, and revisions in statistical methods. An observed fall in registrations alone cannot establish the elimination of the underlying economic activity.

Economic incentives, reform, and substitution

International reforms alter the relative costs and benefits of corporate structures, sometimes without changing the non-fiscal services for which financial centers are used. A reform that raises the cost of an arrangement may lead to consolidation, replacement of intermediaries, changes in funding or changes in substantive activity. The direction and size of the response are empirical questions.

The OECD Action 6 minimum standard and its peer-review framework address treaty shopping. The Pillar Two model rules establish another major international tax-policy framework, accompanied by continuing interpretation and implementation guidance. Neither framework makes a historical network classification a determination of an enterprise's tax obligations. Implementation differs across jurisdictions and over time; this monograph does not claim a complete current legal inventory.

An economic evaluation of anti-abuse measures should distinguish changes in nominal routing from changes in fiscal outcomes. If a firm removes one intermediary but retains the same allocation of productive activity and taxable profit, the visible network changes while the intended fiscal objective may remain unmet. Conversely, a change in profit allocation can improve revenue outcomes even if the corporate graph retains many of its existing entities.

The same concern applies to transparency policy. Disclosures that make ownership links visible can initially increase measured concentration by revealing previously unobserved connections. A rise in a measured indicator after reform may therefore reflect improved information rather than expansion of the underlying activity. Evaluation should assess data coverage and substantive behavior together.

A policy aimed at one hub may induce substitution toward other jurisdictions. Concentration in a few centers can make coordination more administratively tractable, but concentration does not establish that substitutes are absent. Substitution costs depend on institutional capabilities, legal compatibility, funding relationships and the costs of changing organizational arrangements. Without that evidence, predictions of complete displacement or complete extinction of an offshore function are equally unwarranted.

The policy objective should guide the metric. Revenue protection calls for a tax-base and compliance analysis. Financial stability calls for exposures, funding and liquidity analysis. Investment policy calls for ultimate activity and productive outcomes. Corporate transparency calls for entity-level information on ownership and control. A shared map of important centers may inform all these tasks, but no single centrality score can serve as their outcome measure.

This establishes a reason for coordination among national statistical agencies, supervisors and tax authorities without merging their distinct mandates. Statistical transparency supports monitoring; regulatory cooperation addresses shared exposures or cross-border enforcement; adjudication remains governed by applicable legal processes. Each institution should use data for the purpose and within the authority for which it is collected.

Measurement uncertainty and research design

A defensible investigation begins by identifying the estimand. A question about legal organization might use ownership motifs. A question about the geographical recording of capital could use external position statistics. A question about local activity might use employment, revenue generated locally, value added or investment projects. The choice of indicator precedes the choice of dataset.

Sampling and missing information require particular attention. An enterprise dataset can underrepresent companies in jurisdictions with weaker public reporting, or small firms for which commercial information is unavailable. Missing ownership weights can interrupt a path. Missing revenue can affect its relative weight. Because these problems can be related to the same jurisdictional characteristics the study seeks to explain, missingness should not automatically be treated as random.

Several achievable checks would strengthen any future replication. An audit should record the database release and reference date, entity inclusion rules, ownership-edge definition, data cleaning and cycle treatment. Programmatic tests should cover simple paths, parallel paths, same-country segments, round trips and cycles. Weighted and unweighted results should be distinguished. Numerical conservation identities should be tested against the represented motifs. A threshold grid should be chosen before comparing labels rather than tuned to reproduce a preferred list.

A second-stage analysis could compare results across GDP-normalized and absolute measures. Markets change over time, so a denominator's source and date must be compatible with the numerator. Small economies may be sensitive to measurement error or economic shocks in the denominator. Reporting ranges or showing alternative normalizations can make the implications clearer without erasing the baseline result.

A cross-layer comparison would be especially valuable. Observed ownership motifs can be compared with official direct-investment positions, security issuer and parent data, banking residence and nationality statistics, and available aggregate SPE series. The construction of a comparable crosswalk requires respecting differences in stocks, flows, gross and net reporting, consolidation and reference date. It is scientifically preferable to report incompatible observations separately than to force them into one numerical estimate.

For causal identification, institutional changes may offer opportunities for difference-in-differences or event-study methods. Such studies require a defined treatment, credible comparison groups, assessment of pre-treatment trends, attention to simultaneous reforms, and sensitivity to anticipatory reorganization. Network spillovers can violate assumptions of independent treatment because affected activity may migrate to comparison jurisdictions. Empirical design should explicitly investigate such interference.

Qualitative evidence contributes a distinct type of validity. Documents and interviews can help clarify the functions of entities, motives for organization and operational constraints. They cannot alone establish representative aggregate shares. Combining them with statistical evidence can show whether a proposed explanation is consistent with institutional practice. In particular, observed relationships should not be assigned motives without supporting evidence.

The present dossier is a bounded literature synthesis. It uses selected foundational research and official standards, with claim-specific links, and formal illustrative calculations. Its scope is limited by the lack of proprietary microdata and direct access to some full texts. It does not undertake an independent causal study, exhaustive systematic review, current jurisdiction-level reclassification or complete legal implementation survey. Several bibliographic records were accessible only as indexed authoritative abstracts or excerpts; those limitations are recorded in the evidence ledger. Claims resting on such records are restricted to the accessible material.

Integrated economic interpretation

The economic significance of conduit and sink roles lies in the relationship between institutions that connect financial claims and institutions in which ownership-linked value is recorded. The two positions are defined within a specified network and are capable of being occupied simultaneously. Their separation can reveal important differences in the organization of multinational enterprises that are obscured by bilateral aggregates.

From the evidence reviewed, three substantive conclusions follow. First, gross concentration and intermediate organization are jointly important. A sink with a high normalized score and a conduit with high conditional path weight may serve complementary economic functions even when one has a much smaller domestic economy. A policy or statistical analysis concerned with only one of them risks missing relevant links.

Second, the apparent geography of capital is contingent on the observational perspective. The legal residence of a vehicle, ultimate investor, location of operating activity, residence of a banking office and nationality of a financial group can all describe one connected arrangement accurately. The distribution of legal claims and distribution of economic risks may differ. A consistent account preserves those distributions as separate observations and examines the relationships between them.

Third, the relevance of these roles to public revenue, development and stability is conditional on the particular mechanism. Financial routing affects some headline investment measures; ownership topology reveals institutional dependencies; profit allocation influences tax bases; funding and contractual claims govern many channels of financial risk. Each mechanism requires its own evidence. Structural measures can inform investigative priorities but cannot establish the outcomes of these mechanisms by themselves.

An instructive comparison is between a jurisdiction serving a large number of small groups and one serving a few very large enterprises. Both may exhibit the same conduit score. For the first, operational disruption might affect many independent clients. For the second, the observed concentration could reflect the organization of a narrow set of enterprises. Depending on market substitutability and financial exposures, their implications for resilience may be different. The equality of the aggregate measure does not imply equivalence for policy.

The boundaries of the concept also explain why a merged universal offshore index would be problematic. Weighting incompatible financial quantities would introduce an arbitrary value judgment about their importance. A small change in index weights could reverse classifications without new economic evidence. For targeted decisions, reporting separate measures, their definitions and their uncertainty is more informative than synthesizing them into a single headline score whose meaning is unclear.

This is an original analytical interpretation of the reviewed research, with a focus on economic measurement and the allocation of claims and activities. It makes no claim that the underlying concepts, descriptive classifications or network techniques originated in this monograph. Its value is the consistent integration of methods, arithmetic implications and policy distinctions needed to interpret conduit and sink roles in the broader study of offshore economics.

Conclusion

Conduit and sink offshore financial centers describe the organization of a particular aspect of international corporate ownership. Their empirical content comes from observed ownership relationships, defined path weights, comparative economic denominators and explicit classification criteria. The distinction between intermediate and terminal locations makes it possible to study specialization and concentration as relationships among jurisdictions.

The foundational ownership research identifies a historically concentrated structure with major conduits in highly developed economies and diverse sink jurisdictions. In the years following that study, developments in investment statistics and research on securities reallocation have reinforced the importance of distinguishing residence from ultimate economic relationships. Those developments provide complementary evidence, rather than updated estimates of the same ownership-based classification.

The main analytical finding is that the geography of recorded claims, productive activities, corporate control, profit and financial risk must be investigated through their respective measurements. Ownership centrality does not measure cash transfers, and pass-through FDI does not measure shifted profits. A center's financial positions do not establish the value of locally produced services. A country's statistical role does not determine the legal status of particular activities.

Sound empirical work records the date, coverage, institutional definition, direction, denominator and valuation of an indicator. Longitudinal evaluation requires comparable data and attention to both economic changes and changes in visibility or reporting. Economic policy evaluation additionally requires evidence about substantive outcomes and plausible counterfactuals.

The central accomplishment of the conduit/sink distinction is to direct attention toward the connections through which offshore corporate organization is constituted. Used with precise definitions and independent corroboration, it helps identify where economic relationships are recorded, where intermediary services are concentrated, and where further research is required. The analysis of those connections belongs within Offshore Economics as a distinct research object with reciprocal relevance to corporate finance, international macroeconomics and financial geography.

References

  • Garcia-Bernardo, Javier, Jan Fichtner, Frank W. Takes, and Eelke M. Heemskerk. 2017. Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership Network. Scientific Reports 7: 6246. DOI: 10.1038/s41598-017-06322-9. Source

  • Vitali, Stefania, James B. Glattfelder, and Stefano Battiston. 2011. The Network of Global Corporate Control. PLOS ONE 6 (10): e25995. DOI: 10.1371/journal.pone.0025995. Source

  • Zoromé, Ahmed. 2007. Concept of Offshore Financial Centers: In Search of an Operational Definition. IMF Working Paper 07/87. Washington, DC: International Monetary Fund. Source

  • Haberly, Daniel, and Dariusz Wójcik. 2015. Regional Blocks and Imperial Legacies: Mapping the Global Offshore FDI Network. Economic Geography 91 (3): 251–280. DOI: 10.1111/ecge.12078. Source

  • Fichtner, Jan. 2016. The Anatomy of the Cayman Islands Offshore Financial Center: Anglo-America, Japan, and the Role of Hedge Funds. Review of International Political Economy 23 (6): 1034–1063. DOI: 10.1080/09692290.2016.1243143. Source

  • Damgaard, Jannick, Thomas Elkjaer, and Niels Johannesen. 2019. What Is Real and What Is Not in the Global FDI Network? IMF Working Paper 19/274. Washington, DC: International Monetary Fund. DOI: 10.5089/9781513521527.001. Source

  • Coppola, Antonio, Matteo Maggiori, Brent Neiman, and Jesse Schreger. 2021. Redrawing the Map of Global Capital Flows: The Role of Cross-Border Financing and Tax Havens. Quarterly Journal of Economics 136 (3): 1499–1556. DOI: 10.1093/qje/qjab014. Source

  • Tørsløv, Thomas, Ludvig Wier, and Gabriel Zucman. 2023. The Missing Profits of Nations. Review of Economic Studies 90 (3): 1499–1534. DOI: 10.1093/restud/rdac049. Source

  • OECD. 2025. OECD Benchmark Definition of Foreign Direct Investment. Fifth edition. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source

  • OECD. 2025. Special Cases. In OECD Benchmark Definition of Foreign Direct Investment, Fifth Edition, chapter 6. Paris: OECD Publishing. DOI of volume: 10.1787/7f05c0a3-en. Source

  • International Monetary Fund. 2022. IMF Announces the Release of a New Database on Special Purpose Entities (SPEs). Press Release No. 22/53, March 2. Washington, DC: IMF. Source

  • Bank for International Settlements. n.d. Locational Banking Statistics: Overview. BIS Data Portal. Basel: BIS. Accessed for this dossier November 18, 2026. Source

  • Bank for International Settlements. n.d. Consolidated Banking Statistics: Overview. BIS Data Portal. Basel: BIS. Accessed for this dossier November 18, 2026. Source

  • Pogliani, Pamela, and Philip Wooldridge. 2022. Cross-Border Financial Centres. BIS Working Papers No. 1035, July 25. Basel: Bank for International Settlements. Source

  • UN Trade and Development (UNCTAD). 2025. Global Investment Trends Monitor, No. 48. UNCTAD/DIAE/IA/INF/2025/1. Geneva: United Nations. January 20. Source

  • OECD. 2024. Background on the Action 6 Minimum Standard and Peer Review. In Prevention of Tax Treaty Abuse: Sixth Peer Review Report on Treaty Shopping. Paris: OECD Publishing. DOI of volume: 10.1787/36cebf8e-en. Source

  • OECD. n.d. Global Anti-Base Erosion Model Rules (Pillar Two). Paris: OECD. Accessed for this dossier November 18, 2026. Source