RESEARCH DOSSIER
Phantom Foreign Direct Investment
Abstract. Phantom foreign direct investment describes direct investment recorded in corporate entities with limited substantive links to their host economies. This monograph examines how jurisdictionally distributed ownership, intercompany financing and statistical residency generate large recorded positions that differ from the geography of productive enterprise. It reconstructs the relevant accounting relationships, evaluates the foundational empirical estimates and their identification assumptions, distinguishes stocks from flows and immediate from ultimate investors, and analyzes the implications for development, taxation and macrofinancial surveillance. The study combines international statistical standards, original empirical literature, and transparent hypothetical accounting illustrations. Its central conclusion is that residence-based financial statistics, ownership-attributed investment measures and production indicators answer distinct and complementary research questions. Contemporary reporting reforms improve the ability to identify these distinctions; historical estimates and current annual flows retain separate denominators and reference periods. The monograph advances an analytical synthesis of established evidence, with no claim of a new empirical global estimate or an independently replicated dataset.
Keywords: phantom foreign direct investment; offshore economics; special purpose entities; pass-through funding; multinational enterprises; direct investment positions; ultimate investor; international investment statistics; economic integration; financial intermediation
Introduction
The geography of foreign direct investment is simultaneously a geography of financial claims, corporate organization and productive enterprise. These three distributions can move together, but the correspondence depends on the structure of the multinational group. A corporate entity situated between a parent company and an operating affiliate can become the recorded host of a large investment position even when the plant, employees and customers associated with the investment are located elsewhere. The analytical object of this dossier is that divergence and its consequences for the measurement of economic integration.
Phantom foreign direct investment (phantom FDI) denotes direct investment recorded in entities whose substantive connection to the host economy is very limited, as operationalized in the modern empirical literature through special purpose entities and estimates of analogous positions. The expression has a specialized statistical meaning. Its usefulness lies in identifying a category of recorded financial investment that cannot automatically be treated as evidence of local productive capacity. It does not imply that the contracts, liabilities, regulatory obligations or financial assets of the relevant companies are imaginary. Even an entity with few employees can have a consequential balance sheet. Its position can be real in the external accounts and misleading in an inference about production.
The central research questions concern the construction of such positions, their identification in external-sector statistics, the empirical evidence about their scale, and the implications for economic research and policy. The argument develops from statistical units to ownership chains, from ownership chains to measurement, and from measurement to interpretation. Particular attention is paid to the difference between a stock at a reference date and a transaction accumulated during an interval. Historical estimates are presented with their own reference years. This study is a critical synthesis of documentary and empirical research with explicitly hypothetical accounting illustrations. It neither estimates a new global series nor claims to replicate proprietary firm-level data. The evidence cutoff is November 18, 2026.
The statistical object of direct investment
Foreign direct investment is a functional category of external-sector accounts. The fifth Benchmark Definition (OECD, 2025) establishes standards for comparable accounting, with additional statistics designed to disclose the economic purpose and geography of investment. The category records financial relationships in which the investor has significant influence over an enterprise resident in another economy. The definition of direct investment (OECD, 2025, paragraphs 71–77) uses at least ten percent of voting power as the threshold. Control for many multinational enterprise statistics normally entails a greater degree of voting power; the two populations need not coincide.
Residency anchors the recording of cross-border claims. An affiliate incorporated in economy B may be a unit of B even when its management and ownership are closely connected with economy A. A resident unit can own assets, incur liabilities and hold equity in another resident or nonresident unit. The existence of a qualifying direct investment relationship determines which financial claims belong to the FDI category under the applicable rules. Equity includes accumulated earnings retained within the investee as reflected in the position, while relevant intercompany debt represents financing claims rather than ownership percentages. Some debt between affiliated financial intermediaries is subject to separate classification rules, which counsels against describing every loan inside a corporate group as direct investment.
Several quantities must be distinguished. A financial transaction belongs to a period. An investment position is a stock at a reference date. Investment income measures returns over an interval. Capital expenditure measures acquisition or creation of productive assets by an enterprise. Gross fixed capital formation is a national-accounting concept with its own definition and exclusions. An acquired shareholding may transfer the ownership of an existing productive company without adding a single machine. A reinvested earning may enter the external accounts without a corresponding bank transfer. Exchange-rate and market-price changes may alter a position even when financing transactions are zero.
These distinctions matter independently of offshore finance. Even perfectly classified non-SPE investment is a financing and ownership statistic. The interpretation of FDI as a quantity of physical investment requires corroborating evidence about the activity of investee enterprises. An adjustment for phantom positions improves the geographical interpretation of the financial measure; it cannot transform that measure into physical capital formation by definition.
Corporate chains and the economic location of capital
Consider an illustrative group consisting of a parent in economy A, a holding affiliate in B, and an operating manufacturer in C. The parent acquires all the equity of the holding affiliate for 100 monetary units, and B uses those funds to subscribe to equity in the manufacturer. Assume that no other cross-border claims, valuation changes or financing instruments are present. Balance sheets record A's foreign claim on B and B's foreign claim on C. The inward FDI positions of B and C each include 100. A sum of the two inward positions equals 200, although the operating company has received only 100 of equity financing in this simplified example.
B's balance sheet contains a corresponding outward claim and inward financing relationship. The duplication arises from aggregating valid claims across separately resident units. Summing cross-border positions measures the volume of financial relationships; it cannot be interpreted as a count of distinct underlying productive assets. Consolidating the three entities at group level would eliminate the two intercompany equity relationships in the group's consolidated accounts. Such consolidation answers a different question because national external accounts need to preserve claims crossing the boundary of each economy.
Suppose B retained 10 for local administration and invested 90 in C. Its inward equity position would be 100, while the outward equity investment would be 90. The net difference of 10 cannot be equated with productive capital in B. That difference might correspond to bank balances, prepaid expenses, a local office or another asset. Nor can a compiler use the difference as an exact measure of funds transmitted during a given year. It is the opening and closing balance sheets and the associated transactions, other changes and classifications that establish the reconciliation.
The international ownership structure therefore introduces two independent questions. One asks where funds and claims are recorded. The other asks which economies ultimately supply capital or control enterprise decisions and where production occurs. Moving from B to A as the attributed source of C's inward FDI corrects an investor geography but leaves B's recorded liability intact unless the analyst produces a separate adjusted series. Removing B as a phantom host and reassigning C to A are distinct accounting operations. Their results must remain traceable to the unadjusted data.
The same organization can serve lawfully documented purposes. Holding entities may centralize group governance or finance, facilitate acquisition structures, administer shareholdings or isolate risks within a legal structure. The existence of tax incentives is compatible with the presence of these functions. The statistical definition therefore classifies entities and claims by observable criteria; determining the purpose, legal treatment or effect of a particular arrangement requires separate evidence.
Special purpose entities and pass-through funds
Special purpose entities (SPEs) offer a practicable starting point for identifying a portion of the divergence. Their classification turns on institutional and economic characteristics such as residence, ownership patterns, employment, local physical presence and the predominance of cross-border relationships. A common conceptual problem is to assume that the legal name of an entity determines its statistical status. A holding company can employ a substantial workforce and perform group functions; another company with a different legal name may have almost no local activities. Classification should use the prescribed criteria rather than the name.
The OECD (2025) chapter on special cases explains that pass-through funding may also occur within operating affiliates, and recognizes particular treatment for fiscal SPEs. An operating company can hold a large portfolio of equity claims on foreign sister companies at the same time as it conducts genuine local production. Conversely, an SPE might have an investment in a resident affiliate that employs workers locally. Entity-level classifications are useful for compiling comparable statistics, while tracing particular funds can answer finer questions about where investment ends.
The distinction has substantial analytical consequences. An SPE-based measure counts positions classified as belonging to that group of entities. A transaction-level pass-through measure concerns funds that continue through an enterprise to another economy. A corporate ownership measure concerns ownership chains irrespective of an actual cash movement at the time of observation. Treating these as interchangeable invites inconsistent comparisons.
Thresholds are operational devices. A company with six employees and a very large balance sheet may have substantial pass-through claims even if it falls outside a statistical criterion applied to an SPE. A company with few employees and a large asset holding may be indispensable to a financing relationship and provide some local value added. Classification is neither a complete assessment of the economic value of the entity nor evidence about a counterfactual allocation of capital.
A measure designed to describe local production should be validated against local operating characteristics. A measure designed to analyze external financing risk should retain the claims and identify their instrument, maturity and counterpart. A measure designed to study fiscal effects requires data on profit, taxable income and relevant laws. The selection of a measure follows the research question rather than a universal judgment about the desirability of the intermediary.
Immediate and ultimate investment geography
In a bilateral inward FDI table, the immediate investor is the counterpart owner or affiliate identified under the statistical framework. That position is significant for contractual relationships between entities in different economies. The investor controlling the wider group may reside elsewhere. The geographical disaggregation standard (OECD, 2025) introduces a carefully specified ultimate controlling parent and discusses alternative attribution methods. Ultimate parent identification follows controlling links through the corporate structure. The relevant country is the country of residence of the identified unit, which may be a corporation or another eligible institutional unit.
A useful comparison distinguishes the immediate investing economy, ultimate investing economy, immediate host economy and ultimate host economy. They are dimensions of a relationship rather than four mutually exclusive sets of funds. The same enterprise group can have multiple operating affiliates, several immediate investors and alternative methods of assigning ultimate investment. The economy of a controlling parent is not necessarily the nationality of each beneficial shareholder, the residence of the ultimate individual saver, or the economy whose banks supplied credit to the group.
Consider a fictional operating affiliate in C with recorded equity of 100, of which 60 is attributed to a direct investor in B and 40 to a direct investor in D. Suppose the first investor is controlled by an entity in A and the second by an entity in E. A proportional investor attribution can associate 60 with A and 40 with E, assuming those amounts represent the relevant eligible positions and both controlling chains are properly identified. Under an approach assigning the full position to the controlling parent of the investee, A might receive all 100 if it ultimately controls C. These presentations answer related but different questions: attribution of financing interests and attribution of control. Methods must be declared alongside published results.
Attribution can also reveal round-tripping, in which an economy's residents channel investment through a foreign intermediary into their home economy. Foreign residence at the immediate ownership level is compatible with domestic ultimate control. Round-tripping overlaps with phantom investment when the intermediary is a low-substance entity, yet either phenomenon can occur without the other. A group can use a foreign operating subsidiary to invest in its home economy; a phantom intermediary may route investment between two entirely different final economies.
These distinctions explain why a bilateral matrix based solely on immediate counterparts can misrepresent the exposure created by multinational business decisions. They also explain why an ultimate-investor matrix is insufficient for monitoring counterparty credit risk. Both matrices are useful when their definitions are explicit and the bridge from one to the other is transparent.
Stock-flow reconciliation and aggregation
The change of an investment position equals the net transactions affecting that position plus revaluations and other changes. In generic notation, closing position P1 equals opening position P0 plus transactions T, price changes Vp, exchange-rate changes Ve and other changes O. The identity applies to a consistent statistical perimeter; a revision to the perimeter itself must be recorded appropriately. A percentage change in P is not necessarily a percentage change in T, and the sum of several annual flows does not necessarily recover a position.
Take a hypothetical inward equity position with an opening value of 100. New investment of 20, a valuation gain of 15 and exchange-rate effects of minus 5 produce a closing position of 130 if other changes are zero. A headline describing an increase of 30 in the stock cannot support a claim of 30 in new foreign financing. If the enterprise instead received no further financing but its valuation increased by 30, the same closing position would result. The two scenarios have different implications for acquisition activity and external financing.
With a group that owns both resident and foreign units, debt and equity can generate different sensitivity to changes in financial conditions. Debt can mature, be refinanced or be converted to equity. An equity claim can change in value when expected profitability shifts. A reorganization can replace direct ownership of a foreign operating company by an indirect chain. Depending on the transaction and valuation conventions, several recorded relationships may change even while employment and production continue unchanged.
There are also alternative presentations of FDI. An asset/liability presentation groups external claims and obligations according to the balance sheet of resident units. A directional presentation organizes positions by the direction of the underlying investment relationship, incorporating prescribed treatment of reverse investment and transactions between fellow enterprises. Mixing the two approaches in a single world total or comparing one country's inward positions with another's liabilities without reconciliation may introduce artificial differences. Negative net positions and flows need interpretation rather than deletion.
For a host with positive total inward positions T and a comparable SPE component S, a simple residual N = T − S is the non-SPE component. A share S/T is meaningful when numerator and denominator refer to the same date, valuation, perimeter and presentation and the denominator is positive. In cases of netted signed amounts, a ratio can fall outside its conventional range; forcing it to lie between zero and one without reporting the operation may conceal real reporting features. The residual is also only as good as the original classification.
An international aggregate of inward FDI should not be added to the corresponding outward aggregate as if the two were independent amounts of worldwide investment. They are alternative observations of relationships, often subject to bilateral statistical asymmetry. Comparisons of inward and outward positions are valuable for identifying gaps, but the reporting perspectives must be preserved.
Sources and the reconstruction of bilateral positions
The main empirical infrastructure combines macroeconomic datasets and, where accessible, microeconomic information about corporate organization. The IMF's Direct Investment Positions dataset records counterpart-economy positions through the survey formerly known as CDIS. Its equity and debt breakdown helps researchers distinguish instruments. The dataset is a source of observations with specified reporting coverage and conventions; it is not a complete register of global ownership chains.
The IMF SPE database announcement (2022) described the initial release as including 25 economies and 2020 annual data. This is the release's stated coverage, not the number of economies available in 2026. The SPE dataset now provides a dedicated route for obtaining statistics on resident SPE transactions and positions. A credible new empirical project should download the selected vintages and code or archive the extraction metadata, including any suppression or revisions and the exact series identifiers.
A bilateral relationship reported by the recipient economy should ideally be comparable to the outward relationship reported by its investor economy. Differences can reflect valuation, timing, classification, the identification of immediate counterparts, consolidation practices, data confidentiality, currency conversion or incomplete reporting. Damgaard and Elkjaer (2017) organize the problem in Searching for Ultimate Investors around bilateral asymmetries, intermediary entities and ultimate attribution. Mirror observations may help fill a missing entry, but their substitution creates an estimate unless harmonization establishes equivalence.
Suppose an outward observation is 120 and the matching inward observation is 100. Replacing the latter by the former is not innocuous: the source choice changes the host's aggregate by 20 and alters bilateral shares. An arithmetic mean of 110 is likewise a convention rather than a discovered true value. A statistical project can prioritize reported inward figures, distinguish mirror-derived observations and test sensitivity to alternative choices. The defensible approach preserves provenance at the level of each reconstructed cell.
Confidentiality presents a further problem. A suppressed entry can be economically substantial in a small jurisdiction with only a few reporting groups. Coding a suppressed observation as zero biases aggregates and potentially reverses rankings. Missing, suppressed, reported zero, and not applicable are different states. The correct handling depends on the disclosure rules and availability of consistent aggregate margins.
Microdata can improve the identification of ownership chains, but coverage is uneven across corporate forms and jurisdictions. A database of company links may omit firms without public filings, fail to capture changes near the reference date, or use operating revenue and asset valuations that differ from external-sector accounts. Linking macro and micro observations therefore involves an estimation design and documented assumptions rather than a direct replacement of one dataset by the other.
The foundational empirical estimates
The modern empirical reference is Damgaard, Elkjaer and Johannesen's global FDI network study (2024), a journal publication following their 2019 IMF working paper. The two versions represent the development of one research program. They are not independent quantitative confirmations. The published study reconstructs investments into SPEs and non-SPEs and reassigns the latter toward ultimate investor economies using macroeconomic and ownership data. Its Table 3 reports total inward positions of $38.566 trillion, estimated phantom positions of $14.653 trillion, and approximately $7.217 trillion of the phantom component supplied by estimation rather than direct reporting.
The authors' FDI database and code page identifies the historical coverage as about 240 economies during 2009–2017. Neither the historical estimate nor the availability of code means that the same share is established in 2026. The estimate is useful precisely because it describes a defined reference period and a documented measurement problem. A new estimate would require new inputs, revised classifications and a comparable estimation procedure.
The ratio of the two rounded Table 3 totals, 14.653 divided by 38.566, is approximately 38.0%. This arithmetic is the present author's calculation from the published summary values, not an additional empirical estimate. Approximately 49.3% of estimated phantom positions in the table are model-derived (7.217 divided by 14.653). A high coverage rate for unadjusted total FDI thus coexists with substantial dependence on estimation for the classification of its constituents. This distinction is central to assessing the reliability of the adjusted data.
The result also shows why statements about measurement must refer to a denominator. The fraction of worldwide inward direct investment assigned to a class of host entities is not the fraction of global wealth held offshore, the fraction of tax revenue lost, the fraction of cross-border payments passing through a center, or the fraction of enterprises that evade tax. Each of those quantities requires an independent population and measurement design.
Identification, extrapolation, and uncertainty
The 2024 study estimates the share of the non-SPE component where detailed reporting is unavailable, using the relationship between FDI intensity and the non-SPE share observed in a subset of reporting economies. This makes a macroeconomic signal informative about otherwise unobserved classification. The reasoning is plausible: the balance sheet of a small jurisdiction can contain large claims arising from its role as a financial intermediary. The predictor, however, does not identify the function of a particular corporation or the reason a particular investment was carried out.
Consider two hypothetical economies with comparable GDP of 100 and inward FDI positions of 300. One is the host of a highly valuable export enclave, whose operating assets and expected earnings account for substantial foreign ownership. The other specializes in holding companies with foreign claims matched by foreign financing. A predictor based solely on FDI intensity cannot distinguish these cases. Employment, sectoral data, instrument composition and outward positions provide additional information. Even those indicators need to be matched to the statistical perimeter of inward direct investment.
The dependent share and predictor in an intensity model can contain common variables. Total FDI appears in the denominator of the share and the numerator of the intensity measure. Measurement errors in total FDI may therefore induce mechanical covariance. A fitted relationship may also depend on influential observations, especially when a small number of jurisdictions contain extremely large positions. Its out-of-sample performance, sensitivity to definitional changes and uncertainty across economies deserve more attention than an apparently precise fitted global total.
The empirical research reports sensitivity scenarios, including observed SPE positions alone and alternative assumptions for nonreporting centers. Such alternatives illustrate dependence on missing-data assumptions. They should not be converted into a formal confidence interval unless the estimation procedure supplies an appropriate probabilistic interpretation. For a new estimate, uncertainty arising from valuation, sampling or missing-data assumptions should be distinguished from uncertainty about the very category being measured.
A useful diagnostic is to publish each adjusted value together with its evidence class: reported classification, mirror inference, entity linkage, model estimate or unresolved observation. An aggregate adjusted series may remain valuable even when certain bilateral edges are poorly identified. Researchers should resist assigning uniform precision to all rows merely because the dataset has a uniform numerical format.
The growing number of national series with separate SPE information creates opportunities for external validation. The same expansion introduces a comparability challenge: earlier historical estimates may use broader or narrower statistical treatments than newly compiled data. A discontinuity after a new reporting standard can reflect reclassification. To infer a change in offshore behavior, researchers need a bridge between old and new classifications or a stable subset observed consistently across the transition.
The distribution of offshore intermediation
The geographical concentration of intermediary relationships is a distinct feature of the contemporary global financial system. A center can specialize in hosting financial claims while neighboring or more distant economies host production. Some economies have substantial activity in both dimensions. The existence of a high FDI-to-GDP ratio is a prompt for investigating the balance sheet structure, not a sufficient criterion for assigning an entire economy a particular legal or political label.
Haberly and Wójcik's (2015) study of offshore FDI gives an estimate of at least 30% of global stock intermediated through tax havens under their empirical approach. That concept concerns routing and cannot be directly pooled with estimates of the share of FDI hosted by low-substance entities. The studies are relevant to the same broad geographical problem, but different classifications and reference periods can generate different results without contradiction.
Garcia-Bernardo and coauthors (2017) use ownership relationships to distinguish conduit and sink centers. A conduit connects locations along ownership chains; a sink concentrates ownership-attributed value. These roles can be analyzed using network metrics and directed relationships. Corporate ownership networks concern relationships between enterprises and the value associated with them; an FDI network is built from national external-account aggregates. Both representations can reveal concentration, but their edges and their weights have different meanings.
For illustrative purposes, imagine three intermediaries in different economies. The first holds interests in many productive companies worldwide. The second owns the first but has few connections of its own. The third finances several sister companies through loans. Their significance depends on whether the investigator measures the number of links, ownership value, betweenness along paths, direct investment positions or cross-border debt. A ranking of centers by one measure does not establish their rankings by every other measure.
Regional integration, language, established legal institutions, commercial networks, professional service capacity, tax arrangements and historical business ties can all affect routing. These mechanisms can reinforce one another. A legal reform affecting one function of an intermediary may produce substitution to another entity type or jurisdiction. The economic researcher should study the entire chain before asserting that a fall in recorded positions in one jurisdiction represents a worldwide contraction in the corresponding activity.
Jurisdictional concentration may make coordinated statistical improvements particularly valuable. A detailed reporting system in a center with high intermediary balances can clarify a disproportionate share of the world's gross positions. Nonetheless, the more relevant criterion for coverage is the function of the enterprise and the continuity of its links, not the adoption of a static historical list of centers.
Tax incentives, corporate finance, and lawful functions
The corporate decision to use an intermediary reflects expected benefits and costs. Benefits may include reduced coordination expenses, access to established financing markets, legal certainty, contractual convenience, and tax consequences. Costs may include administration, regulatory compliance, financing constraints, reputational exposure and the economic consequences of policy change. A corporate organization exists within an institutional environment whose individual provisions can interact.
An analytical distinction between the location of profits and the location of ownership claims is indispensable. Tørsløv, Wier and Zucman's (2023) Missing Profits of Nations investigates profit shifting and corresponding corrections to economic aggregates. That work addresses the allocation of income. Phantom FDI, by contrast, concerns the location and classification of financial investment positions. A large stock of equity may create channels through which dividends or other income are recorded, but it does not establish the amount of taxable profit shifted.
The distinction can be shown in two hypothetical cases. A holding affiliate may own foreign shares worth 100 while its annual income is 2. Another enterprise may have direct investment equity of 20 and earn 12 in a particular year. Ranking the two by FDI stock gives a different order from ranking them by income. Neither sequence alone establishes taxable income, the applicable tax rate or a tax loss. The researcher requires a fiscal counterfactual and evidence about tax bases, transfer prices, applicable treaties and actual payments to calculate revenue effects.
A claim that tax differences cause larger intermediary positions also demands a causal design. Countries offering useful tax arrangements may have developed complementary legal, financial and administrative institutions. The amount of investment might itself influence policy choices. Observed changes associated with reforms could reflect anticipation, other reforms taking effect simultaneously, foreign valuation shocks or changes in reporting standards. A before-and-after comparison is a starting point for hypothesis development, not proof of a causal effect.
Legality must be assessed at the level of the transaction and applicable institutional rules. Tax evasion, avoidance, aggressive tax planning and legally compliant organizational choices are concepts with different criteria. A statistical inference that an entity has limited substance cannot determine criminal responsibility or the legality of a tax outcome. This dossier analyzes the economic incentives and measurable consequences; detailed treatment of treaty interpretation, anti-abuse rules, criminal liability and beneficial-ownership obligations belongs to the related legal research program.
Financial intermediation can yield local fees, administrative employment, legal services and fiscal receipts even with a limited productive footprint associated with the recorded stock. These effects deserve measurement as genuine local activity. Their existence is compatible with the argument that the gross position substantially overstates the amount of production located there.
Economic integration, gravity, and identification
A standard question in international economics concerns how bilateral economic integration varies with market sizes, distance and other factors. An unadjusted direct investment matrix may give particular prominence to intermediary economies because an immediate-investor edge is often a legal ownership relationship rather than the link between the controlling group and the operating market. The discrepancy adds a source of variation to models of bilateral investment.
The working-paper analysis by Damgaard, Elkjaer and Johannesen (2019) reports an improvement of roughly 25% in the explanatory power of conventional gravity variables after adjustment for phantom FDI and ultimate attribution. This is a reported modeling result, not an increase of 25 percentage points and not a universal result for every sample. Better fit demonstrates that the adjusted geography can be more congruent with those predictors. It does not prove that the adjusted measure captures every aspect of economic integration.
Suppose one productive economy attracts investment from three groups headquartered in equally distant foreign economies. All three groups use the same intermediate company jurisdiction, whose direct position accounts for almost the entire host exposure. A model using the intermediary economy's GDP and distance assigns the same origin to relationships that differ in underlying ownership geography. Reassigning the positions to the ultimate investors introduces previously obscured variation. Whether that variation improves a gravity equation depends on the population, the treatment of minority ownership and the model specification.
There are methodological complications. Many bilateral observations are zero or absent, while logarithmic specifications cannot directly use zeros. Selection into reporting may correlate with the level of financial development. Model-generated outcome variables may have estimation errors correlated with the regressors. In particular, if an adjusted FDI series is constructed using host GDP and the subsequent model explains the adjusted series using host GDP, part of the apparent fit can arise from the construction procedure. An informative robustness test should separate reported from imputed outcomes and use measures of actual affiliate activity whenever available.
If the research question is a causal effect of economic integration on growth, the researcher faces additional endogeneity. Productive investment can respond to expected future demand, infrastructure, institutions and existing productivity. Those same factors affect growth. Mechanically removing intermediary observations does not resolve reverse causality or omitted variables. It improves the measurement of the proposed explanatory variable. Identification of an effect still requires a design suited to the intervention and the data.
Production, labor markets, and development outcomes
The economic importance of direct investment in development depends on the capabilities and activities of the enterprises rather than the size of a recorded ownership claim alone. Casella, Borga and Wacker's (2023) review of multinational production measurement examines the problem of using financing statistics as proxies for production. The distinction is fundamental to any evaluation of investment policy.
Imagine two host economies reporting the same inward FDI position of 500 monetary units. In the first, the position finances an operating group with manufacturing facilities, employees and domestic suppliers. In the second, the position is mainly associated with an entity holding foreign interests. The two economies may have very different local capital expenditure, wages, exports and innovation effects. A development policy rewarding the level of the recorded position alone can favor outcomes that have little relation to its objectives.
Even in the first economy, not every type of FDI generates the same effects. A greenfield project may expand capacity, an acquisition may transfer control of an existing facility, an extension may modernize production, and a corporate restructuring may alter balance sheets. Employment effects depend on technology, the size of the operation, labor conditions and interactions with domestic firms. Effects of an acquisition can be positive, negative or heterogeneous across groups and over time.
Useful complementary variables include foreign-affiliate employment, local compensation, value added, revenues, assets, capital expenditure, research expenditure, exports and domestic supplier relationships. These variables must be matched to the definition of the foreign affiliate population. FDI influence and foreign control may cover different companies. Combining data from the two without harmonizing scope can produce apparent discrepancies even in economies with little offshore intermediation.
A national investment-promotion agency should define the outcome against which success is evaluated. If the economic objective is more productive capacity or skilled employment, it should examine project implementation, capital expenditures, job creation over time and associated fiscal costs. If the objective is improved access to international funding, the relevant outcomes include financing conditions and the resilience of external liabilities. The position of a holding affiliate may be relevant to the second objective without accomplishing the first.
This argument does not warrant excluding financial centers from development analysis. Commercial and professional services associated with cross-border investment can contribute to growth and high-value employment. The question is the magnitude and distribution of that contribution. Measuring actual service output and wages provides a clearer basis than attributing all foreign assets recorded in the jurisdiction to its productive economy.
Balance-sheet risk and macrofinancial surveillance
In financial stability analysis, claims routed through low-substance entities may be highly consequential. A resident intermediary can have large foreign assets and liabilities, limited equity buffers and obligations denominated in multiple currencies. The net external position may appear small because assets and liabilities largely offset, yet the gross positions may be exposed to liquidity, legal or valuation risks. Removing such an entity from all surveillance would discard important information.
Consider an illustrative intermediary with external assets of 100 and external liabilities of 100. Its net external position is zero. If a 20-unit liability is payable immediately while the corresponding assets are illiquid, the entity could face a cash shortfall. If the foreign assets and liabilities are denominated in different currencies, an exchange-rate change may affect its solvency. Contractual guarantees may shift the consequences to a parent company or another group entity. These possibilities cannot be inferred solely from net positions.
A legal chain can distribute responsibilities across multiple economies. Bankruptcy proceedings, contract enforceability and restrictions on fund transfers may determine whether the parent is able to support the intermediary. Accounting consolidation at group level removes intercompany claims from the consolidated balance sheet, but national authorities remain interested in claims across their jurisdictions. Both perspectives are necessary for evaluating cross-border transmission.
In a surveillance dashboard direct investment should be separated by relevant financial instruments, immediate counterparts, currency exposure where available, maturity where meaningful, and sectors. An additional ultimate-parent measure can identify common control. A distinction between SPE and operating entities can assist in interpreting the concentration of exposures. The indicators should be linked rather than substituted for one another.
The appropriate policy response also depends on who bears the risk. A funding arrangement with explicit foreign parent support may have a different consequence for local creditors from an arrangement dependent on domestic guarantees. A high ratio of external positions to GDP may warrant examination, but the probability and domestic impact of distress depend on the connection between the entity and local balance sheets. A country-level ratio is a screening device, not a risk assessment of each entity.
The reporting transition and the interpretation of recent flows
In March 2025 the IMF released new macroeconomic statistical standards including BPM7, alongside the updated System of National Accounts. A standard establishes a compilation framework; it does not ensure that every reporting economy has implemented that framework for the same periods. The transition from earlier manuals requires information about country coverage, revisions, breaks and actual release dates.
The 2025 OECD revision specifically emphasizes investment-purpose breakdowns and improved geographical attribution. It gives researchers greater opportunity to distinguish a corporate restructuring from capacity expansion and to examine gross positions that reflect ownership chains. In the standard and supplementary series chapter, ultimately foreign-owned investors' outward positions are used to help delineate an upper bound for pass-through funds. Such an indicator remains an upper-bound approach rather than a direct observation of the exact amount routed.
UNCTAD's January 2026 Investment Trends Monitor No. 50 initially estimated that global FDI flows in 2025 rose 14%, with growth of about 5% after removing the indicated conduit flows. Its subsequent World Investment Report 2026 revised the headline increase to 6%, reaching approximately $1.6 trillion, and reported a 4% increase excluding conduit flows. The difference between release vintages is analytically informative: estimated annual flows can be materially revised as more data arrive.
The 2026 Investment Trends Monitor No. 51 provides interim evidence on first-half flows, including a more than 70% rise in developed-economy inward flows relative to the average half-year in 2025. That comparison is an interim flow measure with a specified baseline. It cannot be extrapolated to the whole year without an additional forecasting method and a statement of uncertainty.
The conduit adjustment used in current investment monitoring is not a current estimate of the global phantom share established in the 2009–2017 research database. Conduit-related flow adjustments may reflect signed flows into and out of particular jurisdictions. An estimate of the stock held in SPEs at a historical year-end is a different variable, derived from different observations. It would be invalid to divide a 2025 annual global flow by the 2017 estimated phantom stock and interpret the result as a growth rate for phantom investment.
Reporting should preserve both the headline series and the adjusted series, identify the excluded observations and explain the purpose of the adjustment. The comparison is most useful when the reader can reproduce the headline and adjusted totals from a documented vintage. When the source presents a revised outcome, the earlier preliminary result belongs to the history of measurement, not the current headline. This dossier uses the report and monitor findings at their stated release scope, rather than implying a new extraction of national source files.
A reproducible research protocol
A new empirical investigation of phantom FDI should begin by defining the target variable, geographic population, observation period, and intended use. A study of stocks by the immediate investor at year-end has a different data requirement from a study of quarterly flows by ultimate control. The first deliverable should be a data dictionary that distinguishes instruments, reporting perspectives, currency units, valuation dates and observed from derived fields.
The second task is to archive the unadjusted official observations, preserving source identifiers and release dates. Data tables should distinguish zeros from missing or suppressed entries. When confidential data cannot be archived or distributed, preserve permitted aggregate outputs, code and a description of the access restrictions. Different series can share a common publication date while refer to different statistical vintages.
The third task is entity and counterpart classification. Code should link resident SPE information to the corresponding aggregate only if their units and denominators agree. Where direct classifications cannot be reconciled, the discrepancy should remain visible. Missing values should not be silently filled by an assumption derived from the definition of a tax haven. If a model is used, identify training observations, predictors, transformations and out-of-sample validation.
The fourth task is the ownership reconstruction. Retain the original immediate-investor matrix and construct a separate ultimate-attribution matrix. Record the redistribution rule, treatment of uncertain control chains, proportional and controlling-parent approaches, and round-tripping treatment. A row or column total that should be conserved under reassignment can be tested; if the convention changes the eligible base, explain the difference rather than forcing the totals to match.
The fifth task is sensitivity. Compare directly reported classifications with imputed results, alternate training subsets and relevant alternative definitions. Test whether rankings or associations disappear when observations of very large financial centers are excluded. Distinguish the impact of measurement changes from the change of any substantive economic estimate. For econometric work, validate the adjusted financial measure against independently measured production or income outcomes where possible.
The sixth task is interpretation and dissemination. Publish the observed and adjusted series side by side with a concise explanation of their purposes. A table of positions can identify the role of financial claims; an accompanying production table can show the scale of local activity. Risk and fiscal analysis require their own evidence. A documented disagreement among measures is a valuable research result. Reducing everything to one index obscures the economic mechanisms that produced the disagreement.
Finally, reproducibility involves more than runnable code. It entails access to input data, versioned transformations, consistent metadata, checksums and a description of inferential limitations. This dossier checks cited source material and illustrative arithmetic. It does not supply an independently reconstructed bilateral dataset. Its computational examples are generic accounting illustrations and its historical calculations are transparent recomputations from published aggregates.
Integrated economic synthesis
The principal analytical result is that a single FDI statistic cannot carry all the interpretations commonly assigned to foreign investment. Phantom positions reveal this limitation especially clearly, because the location of a recorded claim can be far removed from the activities that motivate or support it. The divergence is an outcome of the organization of legal entities across statistical territories, compounded by incomplete information about their functions and control.
Three separate analytical uses can be articulated within existing economic and statistical concepts. The first is a territorial financial balance sheet, whose purpose is to identify resident and nonresident units' claims and obligations. The second is the geography of economic ownership and financing, which investigates ultimate control, routing and risk allocation. The third is the geography of production and development, which examines output, employment and productive investment. These are ordinary research distinctions; no new named theory or classification system is proposed.
The relation among the three uses is consequential. A new corporate holding structure may increase the number and size of recorded cross-border relationships while preserving ownership of the final operating business. It may relocate some income or liabilities and thereby change the distribution of fiscal and financial risks. Its local production effects may be small, although its administration creates some value added. Distinct empirical indicators are required to establish each of those consequences.
Policies seeking to discourage harmful fiscal practices should be evaluated through their effects on relevant tax bases, compliance and reported income, subject to an appropriate counterfactual. Policies intended to attract productive investment should be evaluated through activities of the enterprises and the welfare effects of induced projects. Policies concerned with external vulnerability need to keep gross balance-sheet claims visible. A change in an adjusted FDI statistic can contribute to each evaluation, but cannot be the sole policy metric.
The evidence also suggests a practical symmetry: researchers should retain the broader external accounts rather than discarding positions described as phantom, and statistical agencies should supply supplementary data rather than expect users to infer production from aggregate FDI. The unadjusted financial record and the analytical transformation are complementary outputs. Their differences disclose where multinational organization departs from a simple direct link between the investing and operating economies.
The limitations of this synthesis concern the uneven availability of entity-level classifications, different publication vintages, imperfect ownership data, and limited access to some publisher pages in the present research session. Bibliographic metadata and indexed abstracts are used only for restricted claims where full text could not be opened. The foundational paper's source copy and selected table and figure have been inspected, but the underlying data and code were not rerun. Consequently, the paper supports a historical assessment of measurement and the methodological interpretation developed here, not a definitive worldwide estimate for 2026.
Conclusion
Phantom foreign direct investment is an established object of empirical research in offshore economics. It describes an important portion of recorded cross-border ownership and financing relationships whose local productive connection is limited. The phenomenon follows from the coexistence of internationally organized enterprise groups and national systems of economic accounting. Its measurement requires disclosure of intermediary entities and, in many cases, the statistical reconstruction of unobserved relationships.
The foundational global study finds that the historical magnitude is economically substantial. The precision and scope of that result must be interpreted in light of the reporting and modeling components. The positions recorded in 2017 cannot be treated as a current estimate of global phantom investment. The subsequent improvement in official statistical standards and dedicated datasets creates the basis for better contemporary research, while transitional classifications and changing vintages create new comparability problems.
The main methodological conclusion is to preserve the distinctions among immediate and ultimate counterparts, stocks and flows, foreign influence and control, and financial claims and productive activities. A reduction in intermediary positions might represent a corporate restructuring, a tax response, a change in valuation, a reporting reclassification or a change in productive investment. Each interpretation requires supporting evidence.
The appropriate scientific use of the concept is neither to erase offshore positions from financial analysis nor to accept them as unqualified indicators of enterprise activity. It is to identify the information contained in the original financial relationships, make the transformations used to analyze those relationships explicit, and assess real economic outcomes through compatible independent evidence. This preserves the utility of FDI statistics while enabling more accurate research into economic integration, development, fiscal externalities and financial stability.
References
Casella, Bruno, Maria Borga, and Konstantin Wacker. 2023. Measuring Multinational Production with Foreign Direct Investment Statistics: Recent Trends, Challenges, and Developments. IMF Working Paper 23/113. Washington, DC: IMF. DOI: 10.5089/9798400243530.001. Source
Damgaard, Jannick, and Thomas Elkjaer. 2017. The Global FDI Network: Searching for Ultimate Investors. IMF Working Paper 17/258. Washington, DC: IMF. DOI: 10.5089/9781484329658.001. Source
Damgaard, Jannick, Thomas Elkjaer, and Niels Johannesen. 2019. What Is Real and What Is Not in the Global FDI Network? IMF Working Paper 19/274. Washington, DC: International Monetary Fund. DOI: 10.5089/9781513521527.001. Source
Damgaard, Jannick, Thomas Elkjaer, and Niels Johannesen. 2024. What Is Real and What Is Not in the Global FDI Network? Journal of International Money and Finance 140, 102971. DOI: 10.1016/j.jimonfin.2023.102971. Source
Garcia-Bernardo, Javier, Jan Fichtner, Frank W. Takes, and Eelke M. Heemskerk. 2017. Uncovering Offshore Financial Centers: Conduits and Sinks in the Global Corporate Ownership Network. Scientific Reports 7, 6246. DOI: 10.1038/s41598-017-06322-9. Source
Haberly, Daniel, and Dariusz Wójcik. 2015. Tax Havens and the Production of Offshore FDI: An Empirical Analysis. Journal of Economic Geography 15(1): 75–101. DOI: 10.1093/jeg/lbu003. Source
International Monetary Fund. 2022. IMF Announces the Release of a New Database on Special Purpose Entities (SPEs). Press Release 22/53, March 2. Source
International Monetary Fund. 2025. IMF and the Statistical Community Release New Global Standards for Macroeconomic Statistics. Press Release 25/72, March 20. Source
International Monetary Fund. n.d. Direct Investment Positions by Counterpart Economy (DIP). IMF Data. Accessed November 18, 2026. Source
International Monetary Fund. n.d. Special Purpose Entities (SPEs). IMF Data. Accessed November 18, 2026. Source
Johannesen, Niels. n.d. FDI Database. Research data and replication code accompanying Damgaard, Elkjaer, and Johannesen (2024). Accessed November 18, 2026. Source
OECD. 2025. Disaggregation by Geography and by Industry. In OECD Benchmark Definition of Foreign Direct Investment, 5th ed. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source
OECD. 2025. Main Concepts and Definitions of Foreign Direct Investment. In OECD Benchmark Definition of Foreign Direct Investment, 5th ed. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source
OECD. 2025. OECD Benchmark Definition of Foreign Direct Investment. 5th ed. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source
OECD. 2025. Special Cases. In OECD Benchmark Definition of Foreign Direct Investment, 5th ed. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source
OECD. 2025. Standard and Supplementary FDI Series. In OECD Benchmark Definition of Foreign Direct Investment, 5th ed. Paris: OECD Publishing. DOI: 10.1787/7f05c0a3-en. Source
Tørsløv, Thomas, Ludvig Wier, and Gabriel Zucman. 2023. The Missing Profits of Nations. Review of Economic Studies 90(3): 1499–1534. DOI: 10.1093/restud/rdac049. Source
UN Trade and Development (UNCTAD). 2026. Global Investment Trends Monitor, No. 50. January 2026. UNCTAD/DIAE/IA/INF/2026/1. Source
UN Trade and Development (UNCTAD). 2026. Global Investment Trends Monitor, No. 51. October 2026. UNCTAD/DIAE/IA/INF/2026/2. Source
UN Trade and Development (UNCTAD). 2026. World Investment Report 2026: International Investment in a Turbulent Era. Geneva: United Nations. Source