RESEARCH DOSSIER
Offshore Financial Networks
Abstract. Offshore financial networks organize cross-border capital, corporate ownership, banking claims, securities, currency funding, and payment services across specialized jurisdictions. This research dossier examines the relations connecting financial intermediaries, groups, investors, borrowers, and financial centers. It integrates published economic research, official statistical conventions, institutional analysis, and clearly specified hypothetical accounting illustrations. The analysis distinguishes immediate financial positions from ultimate investor attribution, location from group nationality, gross intermediation from net assets, and credit exposure from funding and operational dependencies. Historical network research and recent institutional sources show why offshore activity must be interpreted through multiple complementary views. The principal conclusion is that the geographic organization of recorded claims, underlying economic allocation, and financial dependence are related but analytically distinct. A defensible assessment therefore preserves instrument-specific networks and reporting conventions before evaluating concentration, welfare, and shock propagation. The study is a qualitative research synthesis; it does not report a newly estimated global network or introduce a named theory.
Keywords: Offshore finance; financial networks; cross-border financial centers; corporate ownership; international banking; securities; financial geography; intermediation; systemic risk; international investment positions
Scientific Object and Research Design
An offshore financial network is the set of economically consequential cross-border relationships in which financial claims, ownership interests, payment services, guarantees, and intermediation functions are organized through jurisdictions serving participants outside those jurisdictions. The research object is the pattern of connection between those participants and functions. It includes the location of financial statements and the institutional arrangements that make the statements possible, but it also encompasses relationships that remain hidden when accounts are aggregated by national residence.
The unit of analysis changes with the research question. For an ownership inquiry it can be a legal enterprise; for a liquidity inquiry it can be a bank office, consolidated banking group, or funding market; for securities it may be an issuer, investor, custodian, or fund. A jurisdiction is the geographic aggregation of these units. The choice of node determines the interpretation of the resulting graph. A national network of holdings may show two highly connected economies even though one of the individual institutions in that connection is the subsidiary of an institution headquartered elsewhere.
The central questions are how different layers of financial activity become connected, why intermediation is geographically concentrated, how those links are observed in statistics, and when they transmit or redistribute risks. The research proceeds through historical and comparative institutional analysis, reconciliation of statistical definitions, and interpretation of published network research. Numerical examples are deliberately fictional accounting exercises. They demonstrate properties of measurement; they are not estimates of existing economies. The study does not construct a comprehensive observed global network, and its empirical scope is the dates and populations of the cited work.
A distinction between financial center classification and financial networks structures the dossier. A center classification asks which jurisdictions meet a given set of criteria. A network inquiry asks how a function performed in one jurisdiction is linked to functions and participants in other ones. A center can host multiple classes of links, and a jurisdiction that fails a particular center threshold can nevertheless provide essential connectivity. The other offshore dossiers may investigate individual vehicle types, instruments, statistical conventions, or regulatory regimes; the present work integrates their relational economic consequences.
Historical Formation and Institutional Economics
International finance developed through the interaction of monetary systems, commercial demand, legal organization, and institutional competition. The IMF background paper (2000) traces the rise of offshore interbank business in the 1960s and 1970s partly to differences in domestic regulatory and tax conditions. Currency deposits and loans could be organized beyond the jurisdiction issuing the currency. A bank's ability to conduct business outside that jurisdiction changed the geography of credit without changing the unit in which the claim was denominated.
This historical separation is economically important. If currency issuance, commercial incorporation, branch registration, and borrower residence coincided, a country-by-country financial matrix would convey much more of the actual organization of financing. International banking weakened that coincidence. The same currency might circulate across a network of nonresident bank branches and corporate affiliates; external finance could expand faster than the domestic economic output of the places organizing it.
As corporate groups internationalized, specialized affiliates developed functions that could be geographically separated from production: issuing debt, coordinating cash balances, holding ownership rights, managing investment portfolios, or providing services to group members. Each function generates a different demand for legal infrastructure, professional expertise, market access, and supervision. The institutional economics of a center therefore concerns the ability to coordinate specialized activities and the transaction costs incurred across the entire route.
Agglomeration can reinforce initial advantages. A dense market of specialists reduces search and coordination costs, provides continuity of service, and can support complex transactions. Other economic advantages may arise from time zones, language, legal predictability, market infrastructure, or institutional affiliations. These are possible mechanisms, not a causal decomposition estimated in this work. Selection matters: successful centers may attract more skilled professionals because they already have substantial business, making a simple correlation between professional capacity and financial scale inconclusive.
Institutional differences can also create regulatory or tax incentives. Economic analysis of those incentives requires specifying which participants face which obligations, which underlying activities they perform, and how their organizational choices affect costs. An aggregate network edge cannot establish anyone's legal compliance, tax motive, or criminal conduct. Legal judgments and individual investigations require independent evidence.
Financial Geography and Functional Specialization
In their study of small international financial centers, Lane and Milesi-Ferretti (2010) identify a varied range of international financial activities. The important implication for a network analysis is that hosting a fund, recording a bank liability, and administering an ownership structure are distinct economic functions. A jurisdiction's total external balance sheet may be large because several such activities are colocated, or because one activity has grown exceptionally large.
Pogliani and Wooldridge (2022) study cross-border financial centers as jurisdictions specialized in international financial activity. Their approach makes specialization relevant for statistical analysis rather than treating offshore as an immutable legal attribute. The related study of their outsize role by Pogliani, von Peter, and Wooldridge (2022) reports that the small economies examined increased their share of worldwide external assets and liabilities from approximately 15 percent in the late 1980s to 30 percent in the late 2010s, while their share of global GDP stayed below 3 percent. The reported result describes a historical group of economies and a gross balance-sheet measure, not the share of world income or concealed wealth.
The financial centers analysis of von Peter and Wooldridge (2026) distinguishes centers specialized in cross-border services from major hubs of international financial decision making. These categories can overlap. The distinction helps explain why a small place with substantial externally recorded assets may rank differently from a large market in which important investment decisions take place.
Geographical specialization can be represented through three separate attributes: scale, intensity, and dependence. Scale measures absolute positions or activity; intensity compares financial activity with a domestic denominator; dependence estimates the extent to which another participant relies on the center for a specified service. A jurisdiction can have large intensity because its domestic denominator is small. It can have large scale because it is a major market. Neither observation alone determines dependence, which requires evidence about partners, alternatives, and switching costs.
The choice of denominator is substantive. External financial assets divided by GDP compares a stock with an annual income flow, yielding a useful measure of relative scale rather than a probability or rate of return. Dividing by domestic financial assets describes the composition of the financial sector. Dividing by all observed global assets measures share of observed scale. A defensible comparative study reports more than one ratio and identifies the year, coverage, exchange-rate convention, and treatment of missing observations.
The relative importance of a financial hub can differ across a bank credit layer, a securities issuance layer, a fund layer, and a payment network. A single global ranking would impose weights across these economically different functions. The analyst would need to explain why those weights answer the question being asked, rather than presenting an index as an objective ordering of all offshore finance.
Network Representation and Interpretation
A directed weighted graph consists of nodes and edges. Let W(i,j) denote a nonnegative weight attached to a directed relationship from node i to node j. Its meaning might be the market value of assets held by residents of i against issuers resident in j, or a banking claim of a group in i on a borrower in j. Each matrix must state its direction convention. Reversing the direction of a bond-ownership edge to represent payment obligations changes the economic meaning of inflow and outflow.
A multiplex description uses a different matrix for each layer and reference date. Ownership shares, deposits, securities, derivatives, guarantees, and payment-message relationships differ in their units, risk properties, and contractual status. A sum of edges across these layers normally lacks a consistent unit of economic interpretation even when some weights are expressed in a common currency. Their interactions can be studied through additional links connecting the same institution across layers.
The out-strength of node i is the sum of W(i,j) over j; its in-strength is the sum of W(j,i). Both are measures of size conditional on the chosen convention. Degree counts observed counterparties, whereas strength measures the aggregate size of positions. A node connected to many small counterparties can have higher degree but lower strength than a node with one very large bilateral exposure. Betweenness measures position on paths under a particular path definition. Its financial interpretation requires evidence that the paths correspond to feasible relationships; ownership control cannot generally be inferred from an arbitrary shortest path.
For a directed graph with n nodes, no self-loops, and m observed edges, density is m divided by n(n-1). Density changes if the sample includes more institutions, the reporting threshold changes, or previously missing links become visible. A comparison across periods therefore needs a stable population or a documented adjustment. Zero values reported by respondents and missing values resulting from confidentiality or nonparticipation must be distinguished.
Suppose a hypothetical center has claims of 60, 30, and 10 monetary units on three counterpart jurisdictions. The normalized shares are 0.6, 0.3, and 0.1. Their squared-share concentration index is 0.46, and its inverse is approximately 2.17. Three equal claims would give an index of one third and an inverse of three. The index identifies unequal exposure sizes; it says nothing about the financial strength of the counterparties or whether the exposures can be hedged.
Consider a payment service centrality measure based on shortest paths. If a direct route requires expensive settlement in another currency while a longer route works with existing balances, a purely topological shortest path may be irrelevant. A useful weighted path would need economically meaningful costs, access conditions, and direction. Even then, an observed institutional route may reflect contractual constraints excluded from the model. A graph is an abstraction of a particular set of relations rather than a substitute for institutional analysis.
The empirical global banking network study of Minoiu and Reyes (2013) uses cross-border banking flows for 184 countries over 1978-2010 and identifies the global financial crisis as an unusually large network disturbance. It illustrates the use of topology over time, while the present analysis distinguishes flow connectivity from outstanding exposures. A decline in new lending can change a flow network without immediately eliminating old claims from the stock network.
Ownership Chains and Corporate Control
A corporate ownership network consists of enterprises linked by equity interests or voting rights. Such a network may describe direct financial ownership, corporate control, or both; the analyst needs to distinguish those concepts. A link with a 30 percent stake carries a claim on economic value under certain assumptions, but it does not necessarily give control. Control can also depend on agreements and dispersed shareholdings.
Garcia-Bernardo et al. (2017) construct a firm-level ownership network from more than 98 million firms and 71 million relationships. Their method identifies 24 sink jurisdictions and five dominant conduit jurisdictions in the study: the Netherlands, United Kingdom, Ireland, Singapore, and Switzerland. The sink metric compares net retained chain value with a GDP-normalized benchmark. The conduit metric isolates intermediate positions in chains connected to identified sinks. These are study-specific measures derived from ownership relations, not observations of every cash transaction or a permanent list of jurisdictions with a particular legal status.
The analysis of motifs in ownership chains matters because an intermediate node can facilitate connectivity between entities that do not maintain a direct ownership relationship. If a jurisdiction has many links but those links mainly terminate there, it occupies a different position from a center that commonly connects sources and endpoints. Aggregate bilateral FDI can suggest large exposure without distinguishing these positions. Firm-level chain evidence supplies a different perspective.
A fictional three-enterprise chain illustrates the difference between control and cash-flow rights. A parent owns 80 percent of a holding company, which owns 60 percent of an operating business. Under a simple proportional equity interpretation the parent's indirect economic interest is 48 percent. Both links nevertheless exceed a majority of voting power if equity and voting rights coincide. The parent's ultimate control and economic share are related but distinct. If voting rights diverge from equity ownership, additional documentation is needed.
The fictional chain is also sensitive to scope. Suppose the holding company's balance sheet includes an operating subsidiary and securities unrelated to the group. An algorithm that assigns all holding-company assets to the operating business would misallocate the third-party portfolio. An ownership tracing model requires link-specific assets, applicable consolidation rules, and a procedure for preserving residual unallocated positions.
Cycles and joint control present further problems. If enterprises own each other, repeated application of simple path multiplication can double count the same underlying value. If an ownership database has an unresolved endpoint, reporting it as the ultimate owner can give false certainty. Computation should specify stopping criteria, how loops are handled, how alternative paths are combined, and where unresolved value remains.
The cited corporate study acknowledges missing ownership data and its possible geographic selectivity. The analyst should consequently treat absence from the observed graph as incomplete information when reporting is unequal. It is also possible for ownership chains and actual cash flows to follow different routes. The two cannot be made equivalent merely by attaching a monetary estimate to every ownership edge.
Foreign Direct Investment and Pass-Through Positions
The IMF direct investment positions dataset compiles bilateral direct investment relationships using immediate counterparts. It supports an economic description of legally recorded international investment, including positions involving related enterprises. The OECD's fifth-edition Benchmark Definition (2025) expands the analytical treatment of investment purpose, ultimate host and investor geography, and pass-through funds. The geographic disaggregation chapter calls for separate identification of resident special purpose entities and supplementary inward investment positions by ultimate investing economy.
An immediate counterpart relationship answers a precise question: which institutional counterpart in another economy holds or owes the recorded investment position? An ultimate investor relationship answers a different question involving the control or origin of investment as defined by the relevant statistical framework. Properly compiled immediate positions retain their validity even if analysts subsequently restate the geographic distribution.
The global FDI network estimates of Damgaard, Elkjaer, and Johannesen (2019) place the share of phantom FDI near 40 percent of global FDI in 2017, compared with slightly above 30 percent in 2009. These figures are estimates based on the study's procedures and historical data; they describe a category involving corporate structures without corresponding substantial activity at the recording location. They must not be read as direct measures of illegal finance, the global tax gap, or unproductive economic activity elsewhere in the chain.
A simple accounting illustration makes the distinction concrete. Assume an investor resident in A provides 100 units of equity to a holding company resident in B. The holding company invests the same 100 in equity of an operating enterprise resident in C. B has an inward position of 100 and an outward position of 100. A bilateral matrix records A-to-B and B-to-C positions. The sum of these two edges is 200, and under the assumptions the final operating enterprise receives 100. Both edges are genuine financial positions.
If analysis aims to attribute the operating investment to its ultimate source, the restated connection is A-to-C, 100. The restatement cannot simply be added to the immediate network as another 100 units of investment. The correct treatment depends on whether the objective is legal position measurement, financial intermediation, or final investment attribution.
In the example, the net external position of B attributable to these two investments is zero when both assets and liabilities are valued at 100. B nevertheless supplies a financial service and maintains an intermediate balance sheet. If the operating asset loses value, the value of the holding company's equity may change, depending on accounting, contractual, and financing arrangements. A zero net position at inception does not imply the absence of operational exposure or future economic consequences.
The presence of intermediation and pass-through finance complicates the interpretation of investment-to-output ratios. Economic development depends partly on what capital supports: new capacity, acquisition of existing assets, financing of an enterprise, or organizational changes. A high inward FDI position alone cannot distinguish these outcomes. Separating instrument, purpose, residence, and ultimate investor improves the interpretation without denying the usefulness of the original recorded positions.
Securities Markets, Funds, and Ultimate Exposure
Portfolio securities add layers of issuer and holder geography to the offshore network. The IMF portfolio positions dataset is the data infrastructure for observed holdings across counterpart economies under its survey and compilation framework. Coverage is voluntary, and the interpretation of reported positions must reflect the accounting basis and reporting population.
A resident fund might own securities issued in several economies while investors elsewhere own shares in the fund. The fund domicile connects holder-side liabilities to issuer-side assets. A conventional investment position reports the immediate claim on the fund. A look-through exercise can estimate its underlying economic allocation, provided portfolio composition, investor shares, and valuation dates are available.
Assume households in A buy 100 units of shares in a fund in B. At the reference date, the fund holds securities worth 45 units in C, 35 in D, and 20 in E; ignore leverage, fees, other liabilities, and uninvested balances. A residence-based graph includes A-to-B of 100 and separate B-to-C, B-to-D, and B-to-E links. A simplified proportional look-through assigns A-to-C of 45, A-to-D of 35, and A-to-E of 20. The 100-unit total is preserved in the look-through. Adding both perspectives would confuse intermediation with final allocation.
This example is not sufficient to reconstruct a real leveraged fund. Debt financing changes the relationship between fund net asset value and gross securities holdings. Derivatives can add economic exposure without large initial market value. Separate share classes may have different rights or hedges. An investor's allocation cannot be inferred accurately without the relevant portfolio and contractual information.
The issuer side also changes the map. Coppola et al. (2021) show that recording offshore issuance by affiliate residence can change the interpretation of cross-border financing and its currency composition relative to group-level nationality. Identifying the parent company behind an offshore issuer can illuminate exposure to the broader corporate group. Such a restatement, however, does not extinguish the contractual obligations of the actual issuer.
A creditor holding a security issued by a special-purpose affiliate may possess different rights from a creditor of the parent. Guarantees, priority, local insolvency rules, and whether the parent is obliged to support the subsidiary affect recoveries. The economic distinction between group affiliation and legal recourse should remain visible. Aggregate group-level allocation is useful for assessing common exposure, but it is an incomplete instrument-level loss analysis.
A fund may also use custodians and securities depositories outside both the fund domicile and the investor's residence. The custodian supplies recordkeeping and settlement services. Its geographic location cannot automatically identify the beneficial owner. The investor, issuing unit, fund, custodian, and parent group may each belong to a different jurisdiction. These relationships explain why securities networks require separate geographic dimensions rather than a single assignment of the security to one country.
Banking Networks and the Geography of Credit
The banking layer links depositors, bank offices, banking groups, and borrowers. The locational accounting perspective records offices and their counterparties by residence; the consolidated perspective reorganizes positions by banking group. The BIS consolidated banking statistics describe international claims from the nationality perspective of the parent group and eliminate internal positions within the consolidated perimeter.
A loan through an offshore office can contribute to the external balance sheet of its host economy while being a credit decision managed within a group headquartered elsewhere. Both questions matter. Authorities responsible for an office's local funding may need to understand its immediate liabilities, whereas group supervisors may be more interested in the consolidated exposure. Combining the two datasets without accounting for their different populations and scopes risks double counting.
Suppose a bank group headquartered in A funds a branch in B with 100 units. The branch lends 100 to a borrower in C. A locational exposure matrix records an internal group claim connected to B and the branch's claim on C. A group consolidation removes the internal A-to-B claim while preserving the claim on C. If A itself raised the funding externally, the group's external funding liability also remains. This is a hypothetical schematic, not a literal BIS reporting template.
The credit network can therefore include a jurisdiction with an economically important branch even when the corresponding consolidated matrix does not display that place as the lender. A local shock may affect branch operations or funding; a group shock may affect the branch even when its local borrowers remain sound. A comprehensive assessment benefits from both residence and consolidated perspectives.
Diversification in the credit layer concerns who owes the claims, whether losses are correlated, and how the claims respond to macroeconomic events. Exposure to ten subsidiaries of the same distressed group is unlikely to provide the same diversification as exposure to ten independent counterparties. Straightforward counterparty counts must be supplemented by group identity, sector, currency, maturity, collateral, and guarantees.
An offshore banking center may specialize in serving external customers rather than domestic borrowers. Its contribution to local production depends on compensation paid to employees and suppliers, profits attributable under the relevant accounting framework, and related activity. Gross external credit positions do not directly equal this value added. Similarly, a bank's deposit liabilities represent obligations to its customers, not bank wealth available without corresponding claims and risks.
The Global Currency and Funding Layer
Currency adds another geographic dimension. A loan denominated in US dollars can link two non-US residents. The location of a funding market, the residence of the borrower, the headquarters of the lending group, and the sovereign that issues the currency are distinct. The international dollar funding report of the Committee on the Global Financial System (2020) treats the dollar's international use as central to global funding markets and analyzes the vulnerabilities associated with that structure.
McGuire and von Peter (2009) use banking statistics to reconstruct consolidated positions in their global dollar shortage study. They connect expansion of dollar assets in major non-US banking systems with funding pressure when interbank and swap-market functioning deteriorated during the financial crisis. The important economic mechanism is the mismatch between the ability to acquire liquid financing and the obligations created by assets, liabilities, and contingent commitments.
A simplified example helps separate foreign-exchange risk from liquidity risk. A bank has a long-dated dollar asset of 100 and a short-dated dollar liability of 100. The currency shares match at the initial exchange rate. If the liability matures tomorrow and the loan is not repaid for a year, the bank needs 100 units of dollar liquidity regardless of the currency match in its balance sheet. The bank may face a funding crisis without an initial net open foreign-currency position.
An FX swap can allow the bank to obtain the needed currency temporarily while creating a future payment obligation. Depending on the accounting data, the full principal cash flows may not appear as conventional balance-sheet debt. A network based only on recorded debt stocks may therefore omit a significant contingent payment demand. Conversely, aggregating every derivative notional as a credit exposure can vastly overstate loss risk. Different measures support different questions.
The economic role of offshore markets in funding is to connect participants that demand internationally usable currencies with those supplying short-term or long-term financing. Their size alone does not determine resilience. Funding concentration, rollover frequency, access to liquid assets, collateral terms, and emergency liquidity arrangements influence whether a disturbance remains local or spreads across the network.
Regulatory liquidity analysis must accordingly identify the currency, reporting unit, maturity horizon, and contingent obligations before comparing funding demand and buffers. The relevant institutional ability to obtain emergency financing depends on the entity and jurisdiction concerned. The existence of a central bank facility somewhere in the global system does not imply universal access to that facility.
Payments and Correspondent Banking
A payment network is connected by service relationships, accounts, messaging infrastructure, and settlement mechanisms. Correspondent banking enables financial institutions to access payment services and currencies through other banks. Rice, von Peter, and Boar (2020) investigate the retreat of correspondent banks and its geographic distribution. Their analysis is valuable because payment connectivity can change even as the total amount of finance intermediated across borders remains substantial.
The BIS 2019 CPMI data release describes a reduction of about 20 percent in correspondent relationships across seven years using messaging data covering more than 200 jurisdictions. This is a measure of observed relationships, not a 20 percent decrease in worldwide payment value, deposits, or credit. The distinction is essential: the value handled by remaining relationships can grow while the number of relationships contracts.
In a fictional payment network, a small bank sends settlement instructions through two counterpart banks, each with access to the relevant settlement system. If one relationship closes, the number of direct options falls from two to one. The total value of payments could remain unchanged if the remaining bank absorbs the volume. The smaller institution nevertheless becomes more reliant on one provider. Its service resilience may deteriorate despite unchanged transaction totals.
The example also distinguishes topological access from economic access. A technically connected bank may face higher costs, more demanding compliance procedures, lower transaction limits, or restrictions affecting particular customers. Whether an alternative is viable depends on the applicable laws, contracts, infrastructure, and business incentives. A network diagram showing an available path is not proof of usable payment services.
Operational risk in this layer differs from credit risk. A settlement outage, disruption of messaging services, or failure of an intermediary can delay payment even when underlying firms remain solvent. Payment timing may in turn create liquidity pressure. A separate institutional study would be necessary to quantify the effect of an outage or policy change in a particular corridor.
The distributional concern is that relationship contraction can impose costs asymmetrically. Participants with many substitutes may adapt quickly; those with fewer providers and smaller transaction volumes may have less bargaining power. A study evaluating these effects must observe actual service costs, delays, restrictions, and transaction volumes. The network measures are useful indicators of access but do not replace that evidence.
Cross-Layer Connections and Nonbank Intermediation
The 2026 BIS analysis of offshore activity by Aldasoro and coauthors emphasizes the connection between offshore financing affiliates, banking, and nonbank financial institutions. The same underlying economic activity can generate different recorded cross-border positions: a security issued in one jurisdiction, a related-company position connecting another, and a holding by an institutional investor in a third.
Nonbank financial institutions include heterogeneous business models, and their network links must be interpreted in light of these differences. Investment funds, finance companies, insurance companies, and other intermediaries differ in their liability structures, supervision, and capacity to create or absorb liquidity pressure. Grouping them together can be helpful when establishing macroeconomic scale, but may conceal channels that matter during stress.
The FSB report on nonbank leverage (2025) identifies leverage as a financial stability concern. For network analysis, leverage creates a connection between asset-market movements and funding or collateral demands. A fund facing an increase in required collateral may respond by selling assets, obtaining additional finance, or drawing on liquidity buffers. Each response interacts with a different network layer.
A stylized balance sheet illustrates interconnectedness without estimating real exposures. A fund with assets of 120 and liabilities of 20 has net assets of 100. If the market value of all assets declines by 10 percent while liabilities remain fixed, its assets become 108 and net assets become 88. Investors lose 12 percent of net asset value under these assumptions. With additional short-term financing or contingent collateral requirements, the same initial price change could produce liquidity needs not captured by this static arithmetic.
If several funds own similar securities, they can be exposed to the same market decline even in the absence of direct credit relations between them. Common holdings create a two-step path from investor to asset and asset to other investors. This is a different dependency mechanism from direct interbank lending. A stress test must distinguish bilateral losses from common exposures and the effect of asset sales on prices.
An offshore network containing investment vehicles, banks, and market infrastructure may therefore be vulnerable to concentration in services even when direct financial counterparties appear diversified. Multiple funds may rely on the same custodian or financing provider. Conversely, centers specialized in different services may provide redundancy. Establishing either possibility empirically requires matched entity identifiers, information on service providers, and evidence on substitutability.
Measurement Reconciliation and Statistical Discipline
International financial data provide complementary observations, each with a defined statistical object. The IMF's release of BPM7 on March 20, 2025 advances the standards governing external accounts. That release date does not imply that all reporting economies or all historical series have already converted to the new standard. A time series requires documentation of the actual compilation basis used for each observation.
The balance of payments records transactions over a reference period, while the international investment position records external financial assets and liabilities at a reference date. A position can change because of transactions, exchange-rate movements, other price changes, and other changes in volume. In a simplified example, an external asset position rising from 100 to 115 after net transactions of 5 and valuation changes of 10 does not demonstrate a capital inflow of 15. The accounting components must be reconciled separately.
The bilateral direct investment and portfolio datasets identify different instruments and relationships. The consolidated banking statistics apply a different organizational perimeter. The task of constructing a joint analytical description is to preserve the statistical meaning of each variable and then match the parts that can be matched. A common currency of measurement is necessary for some comparisons but insufficient to justify aggregation.
Judson and Kim's (2026) note on securities position asymmetries addresses differences between US Treasury International Capital data and IMF portfolio position data. Such differences motivate an explicit reconciliation rather than an assumption that two counterpart reports are interchangeable. Custody chains, survey coverage, timing, valuation, classifications, and reporting confidentiality are plausible issues to examine; the particular causes and their relative quantitative importance must be established from the actual reporting documentation.
For bilateral position analysis, define the mirror discrepancy for an instrument as the reporting economy's asset claim less the amount attributed to it in the counterpart's liability measure, after harmonizing date and valuation. This statistic is a diagnostic. It is not an estimate of missing illicit wealth. If the coverage differs, it may be impossible to interpret a difference without new information.
Missing values should remain missing until reporting information supports imputation. Confidential values are not evidence of zero. A model-based imputation should specify its predictors, training set, uncertainty, and how imputed values affect network metrics. Sensitivity to alternative treatments of missing observations is especially important when smaller jurisdictions or specific institutional sectors are selectively absent from the data.
A practical evidence architecture for a network study includes the dataset release date, period, vintage, reporting sector, counterpart sector, instrument, currency, valuation basis, and consolidation perimeter. It also records whether each observation is reported, derived, imputed, or unresolved. This metadata is part of the scientific evidence rather than an ancillary technical detail.
The correct treatment of an observed external position depends on the question. For gross intermediation scale, bilateral immediate positions may be indispensable. For corporate exposure, a nationality-based restatement may be useful. For production or beneficial ownership, further information is needed. Each transformation should preserve the original series and create a separate analytical view so that the transformation remains auditable.
Financial Stability and Shock Propagation
Financial networks can absorb shocks through diversification and distribute losses across multiple balance sheets. They can also transmit losses, restrict liquidity, and concentrate dependence on common services. The balance between these outcomes is an empirical and institutional question.
Acemoglu, Ozdaglar, and Tahbaz-Salehi (2015) show in their theoretical analysis that denser financial connections can improve resilience to sufficiently small adverse shocks while facilitating transmission when shocks are large enough. The result belongs to the assumptions of the model and is not a universal empirical claim about every offshore center. It demonstrates why interconnectedness and vulnerability cannot be equated through a single monotonic index.
Consider a creditor with 100 units of equally senior and unsecured claims on two unrelated institutions. In the fictional case, each claim has an exposure of 50 and a default recovery rate of 40 percent. A default by one causes a loss of 30; both defaults would cause a loss of 60. If the same 100 had been placed with one institution, its default would cause a loss of 60. Diversification changes the distribution of losses while leaving the worst loss under these assumptions unchanged. The relative probabilities depend on default correlation and other excluded factors.
Now assume that one institution's default causes distress among lenders that are also its short-term funding providers. A network model incorporating balance-sheet feedback might generate losses beyond the first 30. The size would depend on maturity, recovery, capitalization, liquidity reserves, other obligations, and regulatory interventions. The initial arithmetic alone cannot establish the cascade.
Direct credit losses, common-asset price effects, funding withdrawals, operational service disruption, and changes in expectations are distinct propagation channels. They may operate simultaneously. An observed correlation in market outcomes is not proof of any specific channel. A credible empirical study should formulate a transmission mechanism and test it against well-defined alternatives.
Offshore centers often serve cross-border counterparties. Consequently a stress originating outside a center can affect its recorded liabilities and assets, and disturbances to the center's services can affect participants elsewhere. Local GDP may offer little guidance as to the scale of financial flows passing through the center. Yet large financial scale relative to GDP does not by itself establish the fiscal burden of a crisis; guarantees, loss allocation, resolution regimes and institutional support must be considered.
Stress testing should distinguish loss-bearing entities from entities that provide operational services. A service provider can be important to continuity without carrying a large amount of market risk. A highly leveraged entity can be important as a source of forced selling without occupying the central position in a payment network. The purpose of a multidimensional description is to make these distinct sources of vulnerability visible.
Economic Welfare, Distribution, and Policy
Financial networks can generate economic benefits through access to capital, diversification, specialization, transaction services, and market integration. The costs may include excessive leverage, reduced transparency, tax-base erosion, public backstop exposure, and asymmetric bargaining power. A welfare assessment requires identifying the affected entities, relevant counterfactual, and the allocation of benefits and costs.
A financing structure involving several jurisdictions can lower the firm's funding cost. Part of the gain may reflect diversification of funding sources or economies of scale; another part may arise from differences in taxation or regulation. The balance cannot be estimated by comparing gross international financial positions with output alone. Evidence on contracts, funding rates, investment alternatives, and applicable obligations is necessary.
A host jurisdiction can receive employment, fees, tax receipts, and demand for professional services while the owners and clients responsible for financial positions reside elsewhere. Its domestic benefits depend on the resources employed and the income accruing within the economy. Large externally recorded positions may generate limited local income relative to their face value, although the service sector can still be important to the local economy.
Distributional analysis must also consider participants excluded from networks. A small institution with few viable payment or funding alternatives may experience greater costs than a globally diversified institution. A reform may alter access in one jurisdiction but redistribute business elsewhere. The correct unit of evaluation is the policy's objective across the affected network, not only the change in the balance sheet of its immediate target.
Policy interventions can change edge weights, create new reporting information, or shift business from one institutional form to another. If a supervisory change reduces one type of leverage, banks and nonbanks may respond differently; the presence of substitution is an empirical question. To evaluate a reform, researchers should compare flows, positions, counterpart composition, service availability, and the relevant outcomes before and after implementation while accounting for contemporaneous shocks.
Transparency is valuable when it improves attribution of economically relevant relationships to the correct participants. It must also be designed in accordance with legitimate confidentiality and privacy requirements. Aggregate research data should avoid unnecessary disclosures of individual identities. Classification as an offshore center is not evidence of misconduct by its residents, clients, or institutions.
Coordination between jurisdictions can improve the coverage of an international financial-risk assessment. Authorities may have access to different subsets of the relevant information, and no national balance sheet alone describes the complete network. Cross-border supervision and statistical cooperation should preserve the distinction between legitimate economic activity, legal disputes, regulatory deficiencies, and adjudicated offenses. This research develops economic interpretations rather than legal findings concerning particular institutions.
Empirical Research Protocol and Identification
A reproducible research design begins by stating what the edges represent and selecting a specific population and interval. A banking exposure graph from the end of a quarter cannot be combined indiscriminately with payment messages accumulated over a year. Sector definitions and reporting populations must be matched or explicitly kept separate. Country or territory identifiers must be harmonized without erasing differences in jurisdictional treatment.
The next step is to preserve source observations and create analytical transformations with documented rules. For group attribution, a matched legal entity identifier or verified ownership relation is preferable to a name-only match. Unresolved records should enter a distinct residual rather than being assigned to a plausible-looking parent. Ownership information should be dated consistently with the security or credit position.
A research exercise may compare three views of an illustrative network: immediate residence claims, consolidated group claims, and look-through allocation. Comparing how the weights and rankings of nodes change across the views can diagnose the extent to which geographic concentration reflects booking rather than ultimate exposures. It does not by itself establish why the differences exist.
The choice of null model also affects inference. A bilateral connection may be large in absolute terms but small relative to the scale of the two economies involved. A baseline based on market size differs from a baseline based on reporting-sector financial assets or distance. A claim of exceptional connectivity needs a declared benchmark. Statistical uncertainty should take account of the dependence of observations connected to the same institutions or counterpart economies.
Causal identification requires more than descriptive topology. An event study, difference-in-differences estimate, or natural experiment may be appropriate when there is a well-defined institutional change, comparable affected and unaffected units, and credible assumptions regarding anticipation and spillovers. Financial networks complicate comparisons because a change affecting one node may alter outcomes in nominally untreated nodes. If spillovers are central to the question, a conventional comparison can misstate the effect.
Robustness analysis should vary the aggregation level, treatment of missing observations, minimum edge threshold, classification of institutional sectors, and valuation basis. Rankings or centrality estimates that change dramatically under reasonable alternative assumptions should be presented with their sensitivity rather than promoted as stable results.
Economic identification of the offshore function requires evidence that the relevant service is actually performed for nonresidents or across jurisdictions. Distinguishing a large domestic financial market from a specialized nonresident service center may require information unavailable in aggregate position datasets. This is a valid research limitation and indicates where supervisory, statistical, or firm-level information would improve the analysis.
No independently estimated contemporary network is claimed by this monograph. The examples establish logical accounting relationships and reveal interpretation problems, while the historical estimates and reported results remain tied to the cited publications. A future empirical application should archive code, data vintages, data-access constraints, and the exact definitions of derived quantities.
Integrated Economic Synthesis
The central analytical conclusion is that offshore financial integration is simultaneously a matter of intermediation, attribution, and dependency. Those three matters can be studied together only if their empirical representations remain distinct. The location of an intermediary may describe an important service relationship while failing to identify the final investor or the entity carrying consolidated exposure.
The first consequence concerns measurement. Reallocating the nationality of a claim can improve estimates of who ultimately bears exposure without decreasing the recorded importance of the jurisdiction through which the original contract was organized. Residence and group-nationality measures answer complementary questions. Removing all intermediate edges from the analysis would erase operational dependencies and funding relationships that are economically real.
The second consequence concerns specialization. A center's importance cannot be described adequately by its external assets or GDP ratio alone. Its network function depends on what it connects and whether its services can be replaced. Numerical measures such as degree, volume, or path centrality are informative only within the underlying relationship and population. Different measures can legitimately produce different rankings.
The third consequence concerns risk. A geographically diverse allocation may remain concentrated in common currencies, financial groups, service providers, or shared asset classes. Conversely, a concentrated location of records does not establish that all risk is concentrated in that jurisdiction. Risk analysis needs to connect counterparty exposures with institutional balance sheets, liquidity constraints, guarantees, and infrastructure.
The fourth consequence concerns policy. A change in the geographic location of recorded finance may represent a change in underlying economic activity, a change in legal organization, or a change in measurement. Policy evaluation should identify which of these mechanisms has changed and what happened to the intended outcome. Observing fewer positions in a jurisdiction does not establish that aggregate risk, avoidance, or costs have fallen.
These conclusions are analytical deductions from the distinctions and examples developed in the preceding chapters. They are not a newly named theoretical system or an empirical claim of originality. Their value lies in identifying the conditions under which the different evidence sources can be interpreted together and the conditions under which their results must remain separate.
Conclusion
Offshore financial networks are an integral part of international economic organization. They connect the demand for capital and financial services with specialized providers distributed across jurisdictions. Their significance arises from the relationships between these participants and the institutional conditions governing the relationships.
The evidence examined here supports a plural measurement approach. Ownership chains reveal corporate relationships; bilateral direct investment and securities data record financial positions; consolidated banking data identify group exposures; and payment arrangements reveal service dependencies. Each view makes certain connections visible while leaving others outside its scope. These distinctions explain why external balance-sheet scale, the geography of final investment, and vulnerability to disruptions can diverge.
The analytical contribution of this dossier is to combine these observations through consistent accounting reasoning and explicit examples. It establishes why gross intermediate positions should be preserved for intermediation analysis but should not be interpreted as additional final capital; why group restatements complement rather than erase residence data; and why operational and funding dependencies require separate evidence in financial stability assessment. The resulting economic conclusions remain conditional on what is measured and on the institutional characteristics of the links.
Future research can increase precision by combining dated entity-level ownership data, instrument positions, currency and maturity information, and evidence on the substitutability of financial services. Data coverage, valuation, and missingness should be documented before interpreting centrality or changes in topology. The disciplined connection of these observations gives research on global offshore finance a sound basis for assessing economic development, market integration, institutional competition, and systemic risk.
References
Acemoglu, Daron, Asuman Ozdaglar, and Alireza Tahbaz-Salehi. 2015. Systemic Risk and Stability in Financial Networks. American Economic Review 105 (2): 564-608. DOI: 10.1257/aer.20130456. Publisher record.
Aldasoro, Inaki, Bryan Hardy, Goetz von Peter, and Philip Wooldridge. 2026. International Finance through the Lens of BIS Statistics: Offshore Activity. BIS Quarterly Review, March. Publisher record.
Bank for International Settlements. 2019. CPMI Publishes New Data on Correspondent Banking Networks Showing 20% Reduction in Relationships over Seven Years. May 27. Publisher record.
Bank for International Settlements. n.d. Consolidated Banking Statistics: Overview. BIS Data Portal. Consulted November 18, 2026. Publisher record.
Committee on the Global Financial System. 2020. US Dollar Funding: An International Perspective. CGFS Papers No. 65. Bank for International Settlements. Publisher record.
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. Publisher record.
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. DOI: 10.5089/9781513521527.001. Publisher record.
Financial Stability Board. 2025. Leverage in Nonbank Financial Intermediation: Final Report. July 9. Basel: FSB. Publisher record.
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. Publisher record.
International Monetary Fund. 2000. Offshore Financial Centers: IMF Background Paper. June 23. Washington, DC: IMF. Publisher record.
International Monetary Fund. 2025. Release of New Standards for Macroeconomic Statistics (BPM7). IMF Policy Paper 2025/023. Washington, DC: IMF. Publisher record.
International Monetary Fund. n.d.-a. Direct Investment Positions by Counterpart Economy (DIP, formerly CDIS). IMF Data Portal. Consulted November 18, 2026. Publisher record.
International Monetary Fund. n.d.-b. Portfolio Investment Positions by Counterpart Economy (PIP, formerly CPIS). IMF Data Portal. Consulted November 18, 2026. Publisher record.
Judson, Ruth, and Nyssa Kim. 2026. Measuring Cross-Border Securities Positions: Explaining Asymmetries between U.S. Treasury TIC and IMF PIP (formerly CPIS) Data. FEDS Notes, February 13. Washington, DC: Board of Governors of the Federal Reserve System. DOI: 10.17016/2380-7172.4004. Publisher record.
Lane, Philip R., and Gian Maria Milesi-Ferretti. 2010. Cross-Border Investment in Small International Financial Centers. IMF Working Paper 10/38. DOI: 10.5089/9781451962833.001. Publisher record.
McGuire, Patrick, and Goetz von Peter. 2009. The US Dollar Shortage in Global Banking and the International Policy Response. BIS Working Papers No. 291. Publisher record.
Minoiu, Camelia, and Javier A. Reyes. 2013. A Network Analysis of Global Banking: 1978-2010. Journal of Financial Stability 9 (2): 168-184. Publisher record.
OECD. 2025. Disaggregation by Geography and by Industry. In OECD Benchmark Definition of Foreign Direct Investment, fifth edition. Paris: OECD Publishing. Publisher record.
OECD. 2025. OECD Benchmark Definition of Foreign Direct Investment. Fifth edition. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Publisher record.
Pogliani, Pamela, and Philip Wooldridge. 2022. Cross-Border Financial Centres. BIS Working Papers No. 1035. Bank for International Settlements. Publisher record.
Pogliani, Pamela, Goetz von Peter, and Philip Wooldridge. 2022. The Outsize Role of Cross-Border Financial Centres. BIS Quarterly Review, June. Publisher record.
Rice, Tara, Goetz von Peter, and Codruta Boar. 2020. On the Global Retreat of Correspondent Banks. BIS Quarterly Review, March, 37-52. Publisher record.
von Peter, Goetz, and Philip Wooldridge. 2026. Financial Centres and Offshore Activity. BIS Quarterly Review, March, accompanying box. Publisher record.