RESEARCH DOSSIER

Special Purpose Entities in International Statistics

Abstract. Special purpose entities are a distinct problem in international economic measurement because the location of a financial claim, its ultimate controlling owner, and the economic activity it finances may differ. This monograph examines institutional units, residence, foreign control, the operational definition of resident SPEs, balance-of-payments and international investment-position reporting, foreign direct investment presentations, pass-through financing, ownership attribution, portfolio securities, and the interpretation of national and global aggregates. It integrates primary statistical guidance, national compiler documentation, and published research. Constructed accounting examples demonstrate how additional cross-border ownership links can alter measured global investment without altering final productive expenditure, how gross exposures can coexist with low net positions, and why excluding SPEs cannot completely identify pass-through capital. The empirical discussion distinguishes historical estimates from current observations, statistical classifications from determinations of legal compliance, and source-based interpretation from independent replication. The study concludes that residence-based records and supplemental ownership, institutional, and activity breakdowns provide complementary evidence about the international financial system.

Keywords: Special purpose entities; external sector statistics; foreign direct investment; international investment position; balance of payments; offshore economics; pass-through funds; institutional units; ultimate investor; portfolio investment; statistical harmonization; multinational enterprises

Introduction

A multinational enterprise can organize its factories, borrowers, creditors, shareholders, intellectual property, and treasury functions across different economic territories. Each transaction forms a relationship between specific institutional units. International statistics describe those relationships in a common accounting language, but the geographic pattern of recognized claims need not coincide with the geography of employment, production, final control, or the ultimate exposure to loss. Special purpose entities (SPEs) make that divergence especially visible.

A resident entity can hold foreign shares and loans worth many times the value of goods and services produced in its jurisdiction. The associated claims remain part of the financial structure of the world economy. A statistic designed to answer who owes whom can therefore be correct even when it gives an inadequate answer to a question about where capital is productively employed. Treating the two questions as equivalent is a category error.

The principal argument of this dossier is that SPE statistics improve international economic measurement when the original residence-based record is preserved and supplemented by consistent information about institutional character, ownership, instruments, and the destination of funds. Suppressing cross-border positions in the name of economic substance would destroy information about actual claims. Treating every cross-border claim as a local productive investment would assign economic activity to the wrong place. The analytical task is to reconcile the two descriptions and understand the limits of each.

The historical background begins with the expansion of enterprise groups across borders and the progressive adaptation of direct investment statistics to complex ownership chains. The fourth Benchmark Definition (OECD, 2009) expanded the analytical use of direct investment measurement. Its successor belongs to a coordinated revision of the international macroeconomic statistical system. The changes concern both the population of entities and the additional information users require about them. This study examines the resulting measurement problems rather than the legal permissibility of particular corporate structures.

The scope is international external-sector statistics, with particular attention to direct investment, portfolio securities, and the relationship between financial accounts and domestic economic activity. The monograph uses comparative documentary analysis and accounting examples. It draws empirical implications from published research and compiler documentation, not from a new proprietary dataset or a newly estimated causal model. The numbers in the accounting illustrations are invented solely to expose the logic of the classifications; observed numbers are attributed to their sources.

The Unit of Observation

The institutional unit is the elementary building block of macroeconomic accounting. In principle, such a unit can own resources, undertake obligations, and transact in its own capacity. A statistical reporting enterprise, a legal corporation, a consolidated multinational group, and a productive establishment may describe different layers of the same organization. Their boundaries matter because a claim disappears on consolidation only if both parties are inside the chosen statistical perimeter.

The FDI concepts and definitions (OECD, 2025b) explain the relation between institutional units, sectors, investment relationships, and external accounts. A direct investment relationship exists when the relevant cross-border voting-power threshold is reached; influence and control are distinct. In conventional FDI measurement the threshold for influence is 10 percent of voting power, while control requires more than half. A parent with a minority stake can therefore be a direct investor without controlling the investee. This distinction is essential because the international SPE definition requires foreign control, not merely foreign investment.

A corporate group prepares consolidated financial statements for the group as an economic organization. National statistics, by contrast, require an accounting boundary defined by the economic territory. The assets and liabilities of two resident units are domestic relationships even if one ultimately belongs to a foreign parent. The same contractual relationship crosses the external boundary if the counterparty is a separate institutional unit resident elsewhere. A global consolidated balance sheet cancels claims between affiliates that the external accounts of each jurisdiction must retain.

Consider a business group whose parent resides in A and whose subsidiary in B lends to an operating affiliate in C. A global group balance sheet eliminates the two internal links. B's external assets and liabilities nevertheless record relationships between units across the borders of B. A's national accounts and C's national accounts each describe their own sides of different positions. No contradiction arises: three perimeters answer three different questions.

The practical implication is that entity registries need more than a list of company names. They need reliable linkages between legal units, institutional units, reporting groups, residence, and financial instruments. A legal identifier can assist in matching records, but a stable identifier has no economic meaning without these time-varying attributes. Changes in group control, entity mergers, or the relocation of a registered seat can change the statistical classification while leaving a familiar company name intact.

In particular, an SPE flag must be attached to the correct statistical unit. Identifying a passive shell belonging to a resident parent does not by itself establish a separate unit. Conversely, the small local presence of a nonresident-controlled subsidiary need not justify absorbing it into a foreign parent's accounts. A reliable external accounts system starts with a perimeter decision, and only then asks what type of unit lies within that perimeter.

Residence and the Cross-Border Exception

Residence allocates mutually exclusive economic territories to the institutional units that participate in the accounts. It is a macroeconomic concept, separate from nationality, citizenship, the legal domicile of an investor, tax-residence determinations, and the classification of an issuer by nationality for risk analysis. Differences between these concepts are economically informative, provided they are not silently conflated.

For low-substance units with no ordinary physical center of activity, the recognized territory of establishment becomes especially important. The treatment of special cases (OECD, 2025a, paragraphs 442–443) explains why a passive affiliate resident abroad can be recognized separately while a similarly passive entity in the same territory as its parent may require consolidation. The controlling relationship and the residence of both parties determine the treatment.

Suppose a company in A incorporates a passive subsidiary in B, with no employees there. B hosts a separately recognized financial holding unit. If the same group establishes a second passive unit in B controlled by the first and lacking independent institutional capacity, the national compiler must determine whether the second is part of the first. Counting both merely because both have registration numbers may inflate the measured reporting population. That population issue is different from the externality of B's claims on A and C.

An economic territory can therefore acquire important financial links without acquiring corresponding factories or workforces. Those links are not necessarily fictitious contractual obligations. Their location follows the institutional geography of the accounting system, while the activity that generated their value may be elsewhere. A country whose statistical population includes many such entities can show a very large balance sheet and a comparatively limited domestic wage bill.

Residence cannot be replaced by a single ultimate-owner convention across all statistical questions. For a debt claim, its direct contractual issuer can determine enforceability, payment jurisdiction, and the relevant line of external-sector reporting. For aggregate economic exposure, a parent group may be more useful. For production analysis, the location of establishments, labor, and capital inputs is indispensable. These dimensions can be linked analytically, but none is a universal substitute for the others.

The cross-border exception also clarifies a frequent misunderstanding of offshore finance. The existence of a financial center does not make every entity registered there an SPE. The entity-level definition must be applied to each unit. A bank with substantial staff and local intermediation, an operating transport enterprise, and an inactive holding vehicle may share a jurisdiction yet require quite different classification. Jurisdictional labels cannot perform the work of unit classification.

Operational Identification of Special Purpose Entities

An operational SPE category is narrower than ordinary commercial uses of the expression special purpose vehicle. Corporate lawyers may describe a project company as a vehicle created for a distinct purpose, even when it employs a substantial workforce and owns power stations or roads. Statistical compilers need a consistent population rather than a description of contractual purpose alone.

The IMF's SPE data-template guidelines (2022, printed pp. 1–3) organize identification around formal establishment as a resident institutional unit, foreign control, specific objectives of establishment, at most five employees, minimal physical activity, and a predominantly external orientation. The objectives can concern financial access, isolation of risk, regulatory or tax advantages, or confidentiality. Identification is a joint assessment; no single balance-sheet ratio establishes every criterion.

The employee threshold is a practical boundary. Its contribution is comparability, not an explanation of why a corporation acts as it does. A group with five workers in a particular entity and another with six can carry out nearly identical transactions, but the statistical populations may differ. An analyst should therefore treat borderline classifications as a potential source of discontinuity rather than as evidence of a discrete transformation of economic activity.

The nonresident control requirement distinguishes externally controlled entities from resident-owned captive companies. It also concerns chains of influence. An immediate legal shareholder and the entity exercising ultimate control can lie in different jurisdictions. Determining control requires attention to voting rights and the structure of ownership; a large loan alone may create dependence but does not automatically confer the statistical control required for classification.

Physical presence concerns the configuration of labor, tangible capital, operations, and services. Outsourced local administration complicates a simple headcount. The office providing services can be a substantial resident business while its clients remain low-substance entities. Distinguishing the client's production from the service provider's production prevents a false allocation of economic activity. The value of local administrative services is not zero just because the client's international balance sheet is enormous.

External orientation must be interpreted in light of instruments and business functions. A financing vehicle can issue negotiable securities whose secondary-market buyers are unknown to it. The absence of complete holder information creates a measurement problem rather than proof that liabilities are domestic. The OECD's handling of external orientation (2025a, paragraphs 448–449) recognizes that the business model matters in such cases.

A defensible compiler workflow would first establish the resident unit, next determine ownership and control, then examine staffing, activity, balance-sheet structure, and purpose. It would record the evidence supporting each judgment and a separate status for unresolved information. It would also periodically re-evaluate the population. This is an application of documentary classification, not a new normative definition of the SPE.

The distinction between an identified SPE, an entity that fails a particular criterion, and an entity whose status is unknown is fundamental. Unknown employment should not be interpreted as zero employment. A company without known parent information should not be presumed foreign-controlled. The statistical population becomes biased if missing data are mechanically assigned to whichever class is convenient.

Economic Functions and Heterogeneity

A single statistical flag can cover entities performing sharply different economic functions. A passive holding unit may own shares in operating subsidiaries. A financing unit may issue debt, make intragroup loans, or centralize part of a corporate treasury. A securitization structure may hold a pool of claims and finance it by selling securities. Other entities may hold intangible assets or manage specified international transactions. In each case the economic significance is linked to the nature of the instrument and the distribution of decision rights and risk.

These structures can serve legitimate financing and organizational purposes. Isolating risk, organizing syndicated funding, and maintaining a legally distinct pool of assets can be useful even where taxation is irrelevant. The presence of a tax-related objective in the definition is therefore an analytical dimension, not a finding of tax abuse. Legal compliance involves additional rules and factual evidence outside the statistical flag.

A holding entity's assets may consist chiefly of shares in foreign subsidiaries. Its liabilities may be predominantly equity issued to a foreign parent. The resulting exposure has a residual character: it changes as the value of the operating subsidiaries changes. A treasury entity whose assets are loans and whose liabilities are debt securities has a different configuration of contractual maturity, currency, refinancing, and counterparty risk. Securitization introduces a further distinction between the performance of a portfolio of underlying claims and the specific legal obligations attached to each issued security.

For domestic economic measurement, the distribution of these functions matters because service purchases, employee compensation, and relevant production can arise in a different territory from the holding of financial assets. A holding entity that outsources management creates demand for services; the size of its share portfolio is an unreliable indicator of the value of those services.

The same caution applies to intellectual property. A legal title, an economically relevant ownership claim, the research workers who created an asset, and the income derived from the asset may be situated in distinct places. An external-sector transaction concerning an intangible asset is not a measurement of the location of its original research activity. The correct asset boundary and the nature of the transaction have to be established before interpretation.

The scope of this study is the international statistical identification of these configurations, not the corporate-law requirements for creating them or the accounting rules governing particular investment contracts. Related specialist disciplines can analyze the enforceability of claims, regulatory capital, or tax consequences. In this monograph those distinctions establish why the SPE label needs a second dimension describing the economic function.

Stocks, Transactions, and Other Changes

External-sector accounts distinguish a position, measured at a point in time, from a transaction during an interval. The opening position, transactions, changes due to exchange rates and prices, and other changes in volume together reconcile to the closing position. The arithmetic is common to financial assets and liabilities even when their economic interpretation differs.

Let P0 denote an opening position, T net transactions, V valuation changes, and O other changes. Then P1 = P0 + T + V + O. This identity imposes no behavioral assumptions. It is an accounting reconciliation.

Suppose an SPE opens the year with foreign equity assets of 100 monetary units. During the year it purchases further shares worth 20, the portfolio appreciates by 15, and there are no other changes. The closing position equals 135. The increase of 35 is not a net investment transaction of 35; only 20 belongs to the financial transaction component. Treating the additional 15 as a flow would distort the account of financing during the year.

Conversely, a decrease in position can result from a sale, a lower market value, exchange-rate movement, write-off, or reclassification. The statistical treatment depends on the event. If a resident entity ceases meeting the SPE criterion because its economic activity expands, its position can move from the SPE population to another resident category even when no asset has been bought or sold. A comparison between the two populations must take such changes into account.

Net international investment position (NIIP) equals external financial assets less external financial liabilities. This aggregate is useful for assessing the external wealth of the reporting economy, whereas the two gross sides provide information that the net number suppresses. A matched holding balance sheet may have nearly zero net position but substantial legal relationships across borders.

Assume an SPE holds 250 in external assets and owes 245 externally, while the rest of the economy holds 100 and owes 80. The complete economy has external assets of 350, liabilities of 325, and NIIP of 25. Without the SPE, assets and liabilities would be 100 and 80, a net figure of 20. The gross changes from exclusion are much larger than the net change of five. If one used the excluded gross data for a question about financial exposure, it would miss claims of considerable size; if one used inclusive FDI to estimate locally employed capital, it could substantially overstate the relevant financing.

The result is symmetric: the fact that gross positions are very large does not establish large net indebtedness, and a near-zero net position does not establish absence of risk. Different counterparties, currencies, legal maturities, and contingencies can produce economically important imbalances. The measurement should preserve their separate dimensions wherever the available information permits.

Direct Investment and the Two Presentations

FDI concerns a lasting influence relationship represented by cross-border financial positions and transactions among qualifying investors and enterprises. It includes equity and relevant debt relationships, subject to the applicable instrument and institutional exclusions. The financial component is distinct from the statistical measurement of an enterprise group's real activities.

Under the asset/liability principle, a resident's claim on a nonresident is classified as an asset, and its obligation to a nonresident as a liability. The directional principle organizes qualifying claims and obligations around the inward and outward investment relationships, including the treatment of reverse investment and relationships between fellow enterprises. OECD (2025c) specifies standard and supplementary series that support both forms of presentation.

The importance of the directional presentation can be demonstrated by a simple example. A resident parent's foreign subsidiary owes the parent 100, while the parent owes the subsidiary 25. For simplicity assume both are eligible FDI debt claims and there is no other investment relationship. The asset/liability presentation reports assets of 100 and liabilities of 25. The directional presentation reports outward direct investment of 75 as the net relationship. The examples of asset and liability incidence and direction of investment perform different analytical functions.

If instead the resident unit is a foreign-owned intermediary, its relationship to its direct investor and its separate investments in foreign enterprises must be classified individually. Whether a loan is reverse investment or lending to a different affiliate depends on the underlying relationship, not simply on the sign of the flow. The nationality of the ultimate parent can also be relevant for allocating positions between fellow enterprises.

Both presentations are necessary for different uses. Asset/liability figures maintain reconciliation with the international investment position and the distribution of gross claims. Directional series help assess the magnitude and direction of investment relationships and their industry and geographic distribution. A ratio or comparison that uses a numerator compiled according to the directional principle and a denominator from the asset/liability system can be misleading even if both figures come from an authoritative data provider.

Reinvested earnings require special care. They can be recorded as an imputed income distribution to the direct investor and a corresponding reinvestment. That accounting transaction is not necessarily a cash payment across a bank account. A holding company can therefore report direct investment income and associated reinvestment as the value generated by foreign affiliates accrues through the ownership structure.

The composition of FDI also matters for output. A purchase of newly issued shares may finance an expansion, refinancing, or the acquisition of existing assets depending on the circumstances. A purchase of an existing share transfers ownership. An increase in financial investment is not by itself proof of an increase in fixed capital formation. For that inference the analyst needs the enterprise's investment expenditure or comparable direct evidence of changes in productive capacity.

The Multijurisdictional Investment Chain

A numerical chain makes the relationship between residence-based positions and productive investment concrete. Assume all entities are separate institutional units and group equity is measured consistently at the same valuation date. In A a parent contributes 100 to the equity of its wholly owned financing and holding entity in B. The B entity contributes the same 100 to its wholly owned enterprise in C, which purchases productive equipment from an unrelated resident seller in C. Ignore bank balances after settlement, operating expenses, exchange rates, and profits to isolate the ownership chain.

Economy External FDI assets External FDI liabilities Net FDI position
A 100 0 100
B 100 100 0
C 0 100 -100
World total 200 200 0

Every link appears once as an external asset and once as an external liability. The global inward direct investment position in this three-economy example is 200; the equipment acquired in C is worth 100. These figures are different because the number of cross-border investment relationships exceeds the number of final uses. The extra link is a recognized financial relationship, not a statistical counting mistake.

If B's entity meets the SPE definition, separate reporting identifies its participation in the chain. Removing B's inward position from a global compilation of inward positions leaves 100 under these assumptions. Removing B's outward position from the corresponding worldwide outward positions also leaves 100, but now the geographic allocation of that outward number is anchored at A rather than B. Interpretation of such filtered data must document the statistical perimeter.

In an ultimate-investor presentation, C's inward position may be attributed to A where A is the identifiable ultimate investor. This reformulates the geographic counterpart of the position in C. It does not mean C's legal equity relationship with B has ceased to exist. The resulting analytical position should coexist with the immediate counterpart presentation.

Now add a wholly owned intermediate unit D between B and C, also resident in a separate territory. Keep the same ultimate contribution and final equipment purchase. Gross inward positions rise to 300 while expenditure in C is unchanged. The exercise isolates the effect of chain length. A global total can increase because groups add intermediaries, even if no new project is financed. Conversely, eliminating an intermediary can reduce gross global positions without reducing the amount of productive capital.

The example also reveals why estimates of global investment should not apply an arbitrary correction based on the number of entities alone. Real groups have a mixture of debt, equity, unrelated investors, multiple projects, and changing valuations. An SPE can contribute only part of its funding to a particular project. The attribution of productive uses cannot be inferred from a count of intermediaries without information on the amounts and links.

The author's analytical conclusion is that the residence-based network and the ultimate-use analysis describe different properties of the same financial organization. A change in one may be evidence of reorganization rather than a change in the other. A sound statistical system makes this distinction observable.

Pass-Through Funds and the Limits of Exclusion

Pass-through financing concerns capital entering a unit or economy and continuing to another point in an investment structure. Some of that activity is conducted through statistical SPEs; other flows use units with substantial domestic activities. A population of entities and a category of flows are different kinds of statistical objects.

OECD (2025d) describes pass-through and geographic disaggregation in paragraphs 521–522. Segregating identified resident SPEs provides a lower-bound-oriented indicator, while examining foreign-owned residents that are themselves direct investors can support an upper-bound-oriented estimate. This is a measurement convention for the relevant population and presentation, not a mathematically guaranteed bracket for every possible flow series.

Consider an operating company in B with 100 workers and a parent in A. It receives 100 in an investment transaction, buys local productive equipment worth 60, and invests the remaining 40 in an enterprise in C. Assume all balances are settled, the amounts refer to the same interval, and there are no other funds. The company is not an SPE by the employment criterion, yet some of its financing passes through B. Excluding only SPEs would leave the full 100 in B's inward series. Treating the whole 100 as pass-through would understate the portion that financed domestic investment.

The transactions and the uses are observable together in the constructed example; official aggregates may lack the matching needed to establish this correspondence. If the 40 invested abroad was financed by retained earnings instead of the new foreign contribution, the same aggregate outward transaction could appear. Timing and fungibility make direct matching difficult where funds are pooled.

Stock-based estimates require similar care. An inward stock of 100 and outward stock of 40 do not automatically identify 40 of original funds that passed through the institution. The positions could refer to different acquisition dates, valuation bases, or financing sources. A compiler can present information that constrains interpretations without claiming to have reconstructed every financing chain.

Upper and lower estimates require precise denominators. One should not divide gross inward positions by a net outward flow or compare a measure for resident SPEs with a broader population defined through foreign control without specifying the distinction. Negative FDI flows can reflect disinvestment or repayment; conventional shares may become unstable or undefined around zero. Presenting values alongside ratios prevents apparent shares from obscuring the sign and magnitude of the underlying series.

The economic implication of this analysis is that excluding a recognizable subset of institutional units can improve a particular indicator of domestic involvement while leaving other pass-through mechanisms intact. No exclusion rule eliminates the need for empirical information on the real activities of multinational enterprises. This is a conclusion about identification, not a proposal for a new statistical standard.

Immediate Investors and Ultimate Attribution

Immediate-counterpart statistics assign financial relationships to the resident unit on the other side of the claim. Ultimate-investor statistics seek the economy or entity that ultimately controls the relevant investment chain. The two approaches provide complementary perspectives on the organization of investment.

OECD (2025d, paragraphs 548–552) explains difficult cases involving purported ultimate controlling parents with little substantive presence. Finding a company at the top of an available ownership database is not always sufficient to establish that the company is the economically relevant ultimate parent. A company further up the chain can be missing because a registry is incomplete or because the analyst cannot reliably link records. In such cases the uncertainty should be visible in the data rather than hidden by forced attribution.

Suppose all known documents identify a resident holding entity in B as the immediate investor in C. The available register ends at B; another database indicates ownership by a group in A without sufficient evidence to establish control. The immediate-country attribution to B can be valid while the ultimate attribution remains unresolved. Reporting the latter as known A would confuse a plausible hypothesis with a verified classification. Reporting B unconditionally as the ultimate source would confuse the limit of available observation with the underlying ownership structure.

Round-tripping is another analytically distinct phenomenon. Investment from resident investors can traverse one or more foreign entities and ultimately return to an enterprise resident in the original economy. The resulting inward relationship is cross-border at the immediate counterpart level even though ultimate control belongs to the home economy. Its existence does not prove a particular tax or regulatory motivation. The financing route and its purpose must be established separately.

Ultimate attribution may also be applied to the host economy of outward investment. A parent in A investing through B into several economies may want statistics showing the final geographic distribution of its outward investment. Such a distribution requires links from intermediate entities to final destinations and a documented method for allocating multiple exposures. It cannot always be obtained simply by reversing a table of ultimate inward investment.

The ultimate controlling parent of a direct investment group is conceptually different from a natural person identified as a beneficial owner under anti-money-laundering law. The applicable tests, thresholds, objectives, and reporting units are different. Economic control, beneficial ownership, issuer nationality, and the residence of shareholders all contribute useful information, but they should retain distinct fields in linked research data.

For immediate exposure to a particular jurisdiction's payment arrangements or counterparties, the residence-based view can remain indispensable. For some analyses of group financing, concentration, or the origin of strategic investment, ultimate attribution may be more appropriate. Research should state the target quantity before selecting the statistical representation.

Portfolio Securities and Financing Vehicles

SPEs are not restricted to direct investment. A financing affiliate may issue bonds to unrelated institutional investors and lend the proceeds to another company in its group. The investors' bond holdings are ordinarily portfolio exposures, whereas an eligible intragroup loan may belong to direct investment. Although the funds originate in one transaction chain, the instruments have different statistical classifications and economic characteristics.

Imagine investors resident in A purchasing 80 of bonds issued by a vehicle in B, and B lending 80 to its parent in C. Assume the bondholders have no direct investment relationship with B, and the loan qualifies as FDI debt. A has portfolio assets of 80 against B. B has portfolio liabilities of 80 and direct investment assets of 80 against C. C has direct investment liabilities of 80 against B. B has a matched net external position in this simplified example, ignoring its ownership equity and expenses.

An analytical group-nationality view of A's bond position might attribute the bond's group affiliation to C. This provides information about the corporate organization behind the security. It does not authorize changing the bond's legal issuer, governing terms, maturity, or guarantees in the source record. The parent may be a formal guarantor or may provide no legally binding guarantee; that difference matters.

Coppola, Maggiori, Neiman, and Schreger (2021) use ultimate-parent securities mapping to restate international portfolio positions, illuminating exposures to corporate groups concealed by offshore issuance in standard geographic tabulations. Their published abstract also identifies implications for the currency composition of liabilities and alternative allocations by sales. This is evidence for the value of alternative geographic views of securities, not a universal SPE census. An issuer in an offshore location need not satisfy every formal SPE criterion, and the study's portfolio reallocation should not be summed with a phantom-FDI estimate.

A separate securitization example illustrates the limits of nationality. A vehicle in B may hold a pool of loans made by many unrelated borrowers in C and D. The issuer location, the sponsor location, the location of underlying borrowers, and the nationality of the investors represent different dependencies. If the intellectual target is the credit risk of the underlying portfolio, tagging the entire security with one sponsor country can conceal important information. If the target is the legal issuance channel, the same simplification can discard the very fact being studied.

Risk measurement also depends on maturity and currency. A vehicle can issue short-term external liabilities and acquire long-term external assets. Even if both are initially valued at 80, refinancing pressures can emerge before the assets mature. Equal gross amounts do not imply equal market values after interest-rate or exchange-rate shocks. This is why a complete analysis of financing vehicles extends beyond a single measure of their net foreign position.

The inference is that residence-to-nationality mappings are useful complements to securities statistics when their scope and linking rules are explicit. The original securities identities and positions remain necessary for analyzing the contractual layer.

Income, Domestic Output, and the Interpretation of Ratios

The cross-border income associated with an SPE can include dividends, interest, and reinvested earnings. Depending on its activity it can also participate in transactions concerning goods, services, or nonfinancial assets. The correct classification depends on the nature of the transaction and the applicable accounts, rather than the ultimate policy question of the analyst.

Domestic gross domestic product measures production within the economy. Gross national income incorporates net primary income with the rest of the world. A large financial position can exist alongside limited domestic value added; a high level of investment income can be associated with income subsequently paid to nonresidents. These distinctions create different interpretations of ratios constructed with financial stocks and measures of output.

Consider a stylized residence with GDP of 50 and gross external assets and liabilities each equal to 300. The assets-to-GDP and liabilities-to-GDP ratios are both 600 percent, while the net external position is zero. If the gross figures are due largely to cross-border treasury activity, interpreting either gross ratio as a direct measure of domestic productive capital is wrong. Equally, interpreting the zero net position as absence of cross-border financial intermediation is wrong.

Suppose this entity receives 12 of interest from abroad and pays 11 of interest abroad. Assume these amounts are recorded as primary income for the example and ignore taxes, fees, and measurement adjustments. The gross flows sum to 23 in absolute terms; the net interest receipt is one. An accounting presentation that reports only the net amount can be useful for one question about income, but it reveals little about the volume of the underlying payment relationships.

Income is not necessarily identical to cash transferred in a given period. Accrual, reinvested earnings, valuation, and repayment of principal each describe different processes. An analyst who equates a recorded dividend with a cash transfer without checking the definitions can confuse these processes. Similarly, comparing gross payment flows with estimated tax revenues requires careful distinction between tax bases and the statistical measures of income.

Domestic contribution should be evaluated using separate indicators such as wages, expenditure on domestic services, employment, and investment where they can be observed consistently. A low contribution relative to extremely large assets can be meaningful without being literally zero. A specialized services economy may derive income and employment from administering structures whose financial assets and liabilities are largely foreign.

These examples establish why comparisons among jurisdictions require a clear choice of denominator. Population, GDP, gross national income, financial assets, and the resident enterprise population describe different dimensions. No universal ratio expresses the complete importance of an offshore business model.

International Reporting Architecture and Statistical Transition

The standard-setting process reflects the difficulty of achieving comparability when national reporting practices and corporate structures vary. A separate collection initiative gives compilers a common classification framework and a way to produce comparable additional series. The aim is to maintain the integrity of the external accounts while creating more informative analytical breakdowns.

The IMF announced its first international SPE database on March 2, 2022. The initial release included 25 participating economies, annual observations for 2020, and some earlier or quarterly information. The number is a dated fact about the inauguration, not a claim about present country participation. The SPE dataset portal provides a continuing public entry point for the relevant transaction and position statistics.

The IMF's technical guidance organizes reporting around selected balance-of-payments and investment-position components. Direct investment is important, but portfolio and other investment also belong to the statistical picture. Not every country can supply all dimensions at once, and distinctions between priority and additional detail are relevant to empirical comparisons.

On March 20, 2025, the IMF announced the release of BPM7, aligned with the 2025 System of National Accounts. Its announcement encouraged national implementation by 2029–2030. Therefore publication of a revised manual must be distinguished from national adoption in a given period. A time series downloaded during the transition may continue to reflect the earlier system, a revised system, or an explicit bridge.

OECD's fifth Benchmark Definition extends the analytical presentation of FDI in the same broad modernization process. The continuity of residence-based positions remains important when comparing historical and revised series. New classifications need to be incorporated through revised reporting, metadata, and reconciliation rather than by retroactively attaching missing distinctions to every historic observation.

This creates three different dates: the release of an international standard, the implementation date of a national methodology, and the observation dates covered by revised series. A source published in 2026 can describe transactions in 2023; a 2024 observation may be revised in 2027 under a later methodology. The timestamp of downloading a series is a fourth date. Empirical projects should retain all four dimensions when available.

A harmonized conceptual definition does not guarantee immediate identical measurement. Legal registers, available data sources, survey coverage, allocation rules, and the resources of statistical agencies differ. A difference between two national shares can reflect the structure of their economies, differences in implementation, or both. Harmonization establishes the conditions for interpretation; it cannot substitute for the examination of metadata.

The European Evidence and National Compilation

The European Central Bank's external accounts methods e-book (2025) documents SPE reporting in the European context and distinguishes national from euro area aggregation. Its country-level identification can be maintained in the aggregate, even where a country's resident SPE is controlled by a resident of another euro area member. An intra-area financial relationship is external for an individual member but internal for the area as a whole.

Take an entity in a euro area member X that has a controlling parent in member Y and an external asset against a company in Z outside the euro area. X may classify the entity as an SPE under the national reporting population. In the euro area financial accounts the internal equity relationship between X and Y cancels across the area's boundary, while the claim on Z remains external. Country-level SPE totals cannot simply be added and interpreted as extra-area external positions without removing intra-area claims.

The ECB also describes operational complications, including uncertain security holder residence and the use of a longer observation period for the employment criterion as a stabilizing measure. A temporary fluctuation in staff can create misleading changes in counts and balances when classifications change instantly. Stability of the reporting population and timeliness of reclassification must be balanced, with the resulting rule documented in metadata rather than inferred by the user.

National authorities may develop broader analytical populations to address particular domestic concerns. De Nederlandsche Bank's conduit-financing terminology distinguishes internationally defined SPEs from Dutch captive financial institutions, broader multinational financial holdings, historical special financial institutions, and still wider conduit concepts. The classifications are related but not interchangeable. A change in the series label over time can accompany a change in the unit population.

The Netherlands supplies a well-documented example of revision effects. DNB's 2024 benchmark revision reports a revised conduit-sector balance sheet of EUR 4,832 billion, equal to 671 percent of GDP in 2016, and EUR 4,567 billion, equal to 512 percent of GDP in 2021. The revised 159-percentage-point decline in the ratio was steeper than the previously measured 137-point decline. These are figures for a broader conduit sector; they are not global SPE-only data.

The figures reveal two independent changes. First, the nominal balance-sheet total declined by EUR 265 billion across the two years, about 5.5 percent relative to the 2016 level. Second, the balance-sheet-to-output ratio fell substantially. Its movement reflects both the numerator and denominator. The revised ratio cannot be interpreted as a 159 percent decline in the balance-sheet value: the unit of the reported difference is percentage points.

DNB also reports that new data and changes in the classification population raised previously estimated levels while revealing a stronger relative contraction. An analyst who combined the old early-year series with the revised late-year series could misstate the change. The appropriate comparison uses one coherent vintage.

This case does not prove that all conduit-sector changes follow one institutional or tax reform. The statistical authority discusses several possible policy influences. A causal study would require a credible counterfactual, entity-level responses or suitable aggregate controls, and a treatment of contemporaneous restructuring and measurement revisions. A descriptive revision alone cannot deliver that identification.

The wider lesson is that even well-developed official statistics can substantively change as coverage improves. Revision is part of the scientific measurement process rather than a reason to discard official statistics. The resulting vintage information should remain available for researchers reconstructing historical claims.

Empirical Reconstruction of Global Financial Networks

Research has pursued the problem of multinational ownership chains using both aggregate bilateral positions and more granular links among firms and security issuers. The methods reflect different populations, different reference periods, and different possible forms of error.

Damgaard, Elkjaer, and Johannesen (2019), in What Is Real and What Is Not, combine macro and micro sources to estimate a network that separates what they term phantom direct investment from substantive investment and reallocates the latter to ultimate investing economies. Their IMF working-paper abstract concludes that the phantom component may approach 40 percent of global FDI and that geographic reconstruction improves the explanatory power of standard gravity variables by about 25 percent. The estimates concern the paper's historical reconstruction, including the 2009–2017 period, and are not estimates of the current world stock.

The word phantom is used there as an empirical analytical description of investment routed through corporate shells. In residence-based accounts, these positions may remain genuine, recorded financial claims. The two statements concern different intended measures. A measure designed to show capital aligned with substantive local activity can legitimately exclude an intermediary that remains necessary in a record of contractual positions.

The research result is particularly important for empirical models of bilateral investment. If positions are assigned to jurisdictions that serve mainly as intermediaries, an explanatory variable such as distance from the operating enterprise can be applied to the wrong pair of territories. Correcting the geographic attribution may therefore improve the fit of economic models. Improved model fit does not itself prove that every reattributed relationship is correct; the precision of the link and the completeness of the underlying data remain important.

Portfolio studies such as Coppola et al. (2021) approach the problem through security-issuer and parent relationships. Their unit of observation differs from the aggregate country-level FDI reconstruction. The common insight is that institutional geography can move the recorded counterpart away from the economically relevant group. The results should be interpreted in their respective measurement systems and not combined as though both papers measured a mutually exclusive share of one universe of funds.

A further concern is selection. Greater coverage of publicly traded firms can facilitate matching issuers to ultimate parents, while private firms or complex ownership vehicles may remain harder to trace. Dataset coverage, missing observations, algorithmic matches, industry differences, and legal restructurings can change estimated network patterns. A successful empirical method should document these dimensions and report sensitivity to plausible alternatives.

The present dossier does not reproduce either research team's estimation dataset or claim to verify every regression. Its empirical use is confined to the published findings confirmed in accessible publisher records and the conceptual consequences that follow from them. A replication project would require the corresponding dataset versions, linking algorithms, treatment of unmatched units, and code.

Data Collection, Confidentiality, and Validation

A separate SPE series requires a reporting population that can be maintained over time, not simply a one-off survey of entities with large balance sheets. The compiler must establish possible respondent units and then obtain the variables needed to determine residence, control, activity, and external exposure.

Possible administrative sources include company registrations, regulatory filings, financial statements, and enterprise-group registers. Direct surveys can supply assets, liabilities, transactions, and details that are absent from registry documents. Instrument-level records can supplement reported positions where negotiable debt makes counterpart identification difficult. The best source varies by the claim under investigation: a legal register may identify a corporation accurately but contain little information on its balance sheet.

The IMF (2022) discusses the role of cooperation among statistical compilers and regulators in its technical guidance (printed pp. 18–22). When several agencies observe different aspects of the same unit, identifier matching and a documented sharing framework can improve the reporting population. The compiler remains responsible for the statistical interpretation of the matched information.

A database architecture should maintain separate time-stamped attributes for raw legal identity, resident institutional unit, immediate parent, ultimate control where verified, entity category, and instrument classification. Historical observations should preserve the classification that applied in the relevant reference period, or explicitly record revisions. Substituting the current parent for the historical parent creates look-ahead bias in longitudinal research.

Reconciliation can be carried out at several levels. At the unit level, positions and changes must satisfy the appropriate stock-flow identities. At a group level, linked units' reports can identify inconsistencies in claims where the relevant definitions and valuation bases coincide. At a national level, SPE and other units should reconcile with the aggregate external accounts for the same reporting classifications. International bilateral comparison provides another diagnostic, although asymmetries can remain because of valuation, timing, reporting gaps, or counterpart allocation.

Confidentiality is particularly important in small reporting populations. A cell that contains one or two dominant institutions can disclose sensitive firm-level information even after names are removed. Disclosure controls may require suppression or aggregation. The user must distinguish a suppressed observation from zero, and the data distributor should preserve the distinction between missing, confidential, not applicable, and observed zero.

For research, selective disclosure complicates comparisons. Subtracting a published subtotal from a total may recreate protected information or mix populations if the subtotals differ. Estimating omitted data requires documented assumptions and should respect disclosure restrictions. Publishing ranges or aggregating categories can preserve analytical usefulness where exact values cannot be disseminated.

Well-designed validation is therefore both accounting and institutional. An internally balanced table can still misclassify its units; a correctly classified sample can still omit large institutions. Conversely, complete capture of legal entities does not guarantee correct allocation of external positions. The scientific value of the statistics depends on addressing the dimensions together.

A Practical Protocol for Comparative Economic Research

A study of SPE statistics should specify its estimand before selecting measures. The estimand may concern gross external assets, net positions, inward directional direct investment in non-SPEs, the geography of ultimate control, the financial exposure of a corporate group, or local productive activity. These quantities are not interchangeable.

Begin with an explicit population and reporting perimeter. Is the economy a country or a consolidated currency area? Does the source classify SPEs according to the harmonized definition, a domestic conduit concept, or an older reporting category? Are figures inclusive of banks, funds, securitization entities, and operating enterprises? The answer must be taken from the relevant series rather than inferred from a headline.

Next establish the measure. Stocks refer to a reference date; flows cover an interval; income is differentiated from financial transactions. Separate assets and liabilities from inward and outward directional data, and retain information about currency and valuation. A time series in current dollars can change as a result of exchange rates even when local-currency positions remain constant.

A transparent comparison may present the inclusive total, the identified SPE subset, and the remainder. Where compatible, let X be a total external asset position, S its resident SPE component, and N the remaining position. Then X = S + N. If X is positive and meaningful as a denominator, S divided by X reports the SPE share of that specific asset series. It does not estimate the share of proceeds ultimately employed elsewhere, and it does not measure illegality. The same operation for net positions or directional flows requires particular care because signs and denominators can make a share analytically unsuitable.

Suppose an economy reports external claims of 400 in one vintage, of which 250 are attributed to its SPE population. The corresponding remainder is 150 and the SPE share of gross external claims is 62.5 percent. If the data also show SPE liabilities of 245, the SPE net claim is five, but that does not imply that the remainder has a net claim of 150. For the latter the corresponding non-SPE liabilities are required.

For temporal comparison, construct a bridge that isolates recorded transactions, price and currency effects, changes in the reporting population, and revisions wherever available. Avoid interpreting a fall in the share of SPEs solely as reduced pass-through financing. A reclassification from SPE to non-SPE may lower that share while leaving the same financing chain intact.

Compare immediate and ultimate geographic matrices when the latter exist. Retain the reclassification metadata, a category for unmatched or unallocated information, and the possibility that an ultimate-host distribution covers only a subset of the immediate series. Test whether changes in results stem from additional observations, alternative assignment rules, or altered statistics.

Finally, examine independent indicators of real activity appropriate to the research question. These may include employment, compensation, turnover, investment expenditure, income, or establishment-level data. If only aggregate financial series are available, qualify conclusions about domestic productive effects accordingly. This approach permits rigorous descriptive research without inventing observations for missing components or asserting that a relationship has been identified causally.

These are research procedures derived from established accounting and reporting distinctions. They are not proposed as a named model or an alternative international standard.

Financial Stability, Policy, and Evidentiary Boundaries

The geographic location of an SPE can be relevant to financial stability because cross-border financing relationships are also channels through which price, liquidity, credit, and legal disruptions are transmitted. A large gross position can coexist with limited domestic employment and little net external wealth. The effect of a shock depends on contracts, balance-sheet characteristics, parent obligations, and the ability of institutions to adjust.

Consider a funding vehicle with external debt of 100 maturing next month and an external loan of 100 maturing in three years. Its initial net external position is zero. If it cannot roll the debt over, the maturity gap creates a financing problem independent of the net position. Suppose the parent is legally obliged to provide funding: then the parent's liquidity may be affected. If the parent has no such obligation, the outcome can differ. A statistical flag does not resolve the legal and behavioral dependencies.

A risk analysis based on the residence of liabilities may assign the issuer to the vehicle's location. A group-level approach may associate the instrument with the parent group, and an underlying exposure approach may trace the assets held by the vehicle. Each can reveal a different vulnerability. The data need to preserve enough information to connect these views rather than presenting one as the sole true geography of risk.

Tax and regulatory policy raise an adjacent issue. A high SPE share can identify a jurisdiction or sector that merits further investigation of its financial structure. It cannot establish that every recorded institution has avoided tax, that a tax base has been lost in a particular country, or that a payment was illicit. Those conclusions require evidence about applicable rules, taxable income, economic ownership, reporting obligations, and the facts of specific transactions.

Policy changes may affect the number of resident SPEs, the composition of financial instruments, and the activities of entities near the classification threshold. Changes can also encourage the consolidation, relocation, or conversion of existing vehicles with little change in the group's underlying real assets. Observational comparisons around a reform date can confound these responses with simultaneous macroeconomic shifts and improvements in reporting. A study of regulatory impacts requires an identification strategy that distinguishes these mechanisms.

Statistical transparency contributes to policy by making the relevant differences visible. It is not an alternative to financial supervision, corporate reporting, or tax administration. Confidential data can support authorities that have appropriate legal access, while carefully aggregated public statistics enable analysis of system-wide structures. The measurement target should determine the required level of disclosure.

For development policy, the distinction between the location of FDI and the location of employment or productive expenditure is central. An increase in inward positions can occur without a new economic project, while productive capacity can expand with domestic financing rather than a contemporaneous rise in FDI. The absence of an SPE flag is not proof that inward investment generated local jobs. The positive economic effects of international investment must therefore be studied with direct measures of the relevant outcomes.

The scientific position of this dossier is that statistical observability and legal classification complement one another. Evidence of intermediated financial relationships supports an account of the economic system; the interpretation of motives, compliance, and effects must be supported independently.

Integrated Analytical Synthesis

Three forms of geographic information recur throughout the research. The territory of the resident institutional unit determines the original statistical record. The ownership and control chain informs ultimate-group attribution. The territory where resources are deployed describes productive activity or underlying economic exposure. These are ordinary analytical distinctions within existing macroeconomic and financial statistics, not new categories of economic ontology.

An SPE may occupy an important position in the first geography while being peripheral in the third. An operating affiliate may have a substantial position in all three. A financing affiliate may be geographically important in a risk network even when its local production is modest. These observations place a precise limit on inferences drawn from any one dataset.

The first conclusion concerns additivity. The original residence-based assets and liabilities must reconcile at the selected perimeter. Where a supplemental SPE series forms a compatible subset, it can be separated from the total without destroying its traceability. Ultimate-investor or issuer-group reallocations require their own explicit correspondence rules. Combining original and reallocated series as though they were independent exposures creates duplicate counting.

The second concerns the measurement of pass-through activity. An SPE is a classification of a statistical unit, whereas pass-through financing concerns patterns of movement and holdings. The former facilitates measurement of the latter but does not completely identify it. The distinction explains why a narrow exclusion can improve a domestic investment measure and still leave substantial financing routed through operating entities.

The third concerns the interpretation of global totals. A long chain can have more recognized cross-border investment links than a short chain with the same final use of funds. Its measured financial scale can therefore change independently of investment in productive assets. That possibility has implications for comparisons of globalization over time, the geographic allocation of bilateral investment positions, and the interpretation of aggregate FDI-to-output ratios.

The fourth concerns empirical identification. Published reconstructions of direct investment and portfolio securities demonstrate the analytical importance of ownership information. Their findings concern specific data, methods, dates, and populations. They support the value of alternate views rather than the attribution of a single fixed percentage of contemporary world finance to low-substance structures.

The fifth concerns the transition of international statistical standards. Enhanced identification and revised supplementary series strengthen the analysis when implementation status and source vintages are recorded. An apparent break caused by a new definition is itself evidence about measurement, not automatically evidence about economic behavior.

Together these conclusions support a layered and auditable interpretation of existing statistics. An observer should be able to follow a financial position from its original resident unit to the available ownership or exposure information, and return to the original aggregate without losing the position's type, observation date, or valuation basis. The author's contribution is this integrated argument and the worked accounting illustrations. It does not establish a new named statistical system or claim priority for the underlying concepts.

Conclusion

Special purpose entities reveal the institutional architecture of international finance. Their measured assets, liabilities, and transactions can be far larger than their direct contribution to local production because the geographical organization of corporate ownership and financing differs from the geographical organization of employment and productive activity.

International statistical standards address this phenomenon by identifying the relevant resident institutional units and presenting their cross-border relationships in a form that can be distinguished from other activity. The residence-based accounts preserve the legal and statistical counterpart of claims. Supplementary SPE breakdowns, pass-through indicators, and ultimate-investor or group-level analyses provide further information about the organization of the same financial system.

The numerical examples show why gross positions, net exposure, recorded investment directions, and final uses of capital can diverge. The empirical literature demonstrates the significance of those differences for international investment research. National compilation experience shows that the boundaries of the reporting population and the vintage of published data are themselves important scientific facts.

The appropriate conclusion is not to discard recorded relationships that lack a direct association with domestic output. The relationships belong to the evidence. The appropriate method is to identify their character, maintain the accounting perimeter, trace the relevant ownership links where possible, and select the statistical representation appropriate to the research question. The quality of international financial measurement depends on that combination of precision and transparency.

The work establishes the scientific scope of a specialized dossier within Offshore Economics. It supplies a documentary synthesis and explicit accounting arguments for the use of SPE statistics while retaining the limits of the source access and empirical methods actually employed. New estimates of global prevalence, causal evaluations of reforms, and comprehensive entity-level reconstruction require dedicated future empirical research. This study makes no claim to have performed those tasks.

References

  • IMF (2022), Special Purpose Entities: Guidelines for a Data Template to Collect Separate Information on Cross-Border Positions and Flows of Resident Special Purpose Entities. Technical Notes and Manuals 2022/006. Washington, DC: International Monetary Fund. DOI: 10.5089/9798400231490.005. Publisher record or full text.

  • OECD (2025a), OECD Benchmark Definition of Foreign Direct Investment (Fifth Edition), Chapter 6, Special Cases. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Publisher record or full text.

  • OECD (2025b), OECD Benchmark Definition of Foreign Direct Investment (Fifth Edition), Chapter 2, Main Concepts and Definitions of Foreign Direct Investment. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Publisher record or full text.

  • OECD (2025c), OECD Benchmark Definition of Foreign Direct Investment (Fifth Edition), Chapter 7, Standard and Supplementary FDI Series. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Publisher record or full text.

  • OECD (2025d), OECD Benchmark Definition of Foreign Direct Investment (Fifth Edition), Chapter 8, Disaggregation by Geography and by Industry. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Publisher record or full text.

  • IMF (2025), IMF and the Statistical Community Release New Global Standards for Macroeconomic Statistics. Press Release No. 25/072, March 20. Washington, DC: International Monetary Fund. Publisher record or full text.

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

  • IMF (n.d.), Special Purpose Entities (SPEs). IMF Data, dataset portal. Washington, DC: International Monetary Fund. Accessed November 18, 2026. Publisher record or full text.

  • European Central Bank (2025), EU Balance of Payments and International Investment Position Statistical Sources and Methods: B.o.p. and I.i.p. E-book, July 2025, section on special purpose entities. Frankfurt am Main: ECB. Publisher record or full text.

  • Damgaard, Jannick, Thomas Elkjaer, and Niels Johannesen (2019), What Is Real and What Is Not in the Global FDI Network? IMF Working Paper 2019/274. Washington, DC: International Monetary Fund. DOI: 10.5089/9781513521527.001. Publisher record or full text.

  • 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. Publisher record or full text.

  • De Nederlandsche Bank (n.d.), The Various Terms Used for Conduit Financing. Amsterdam: DNB. Accessed November 18, 2026. Publisher record or full text.

  • De Nederlandsche Bank (2024), Contraction of Dutch Conduit Sector Was Underestimated. Statistical News, June 24. Amsterdam: DNB. Publisher record or full text.

  • OECD (2009), OECD Benchmark Definition of Foreign Direct Investment 2008: Fourth Edition. Paris: OECD Publishing. DOI: 10.1787/9789264045743-en. Publisher record or full text.