America's net foreign position is the small difference between two very large numbers

At the end of the second quarter of 2026 the United States owed the rest of the world 22.42 trillion dollars on the official measure. How that figure is built matters more than its size. American residents held 46.97 trillion dollars of foreign financial assets; foreigners held 69.39 trillion dollars of American ones; the net international investment position is the difference.¹ The headline number is about a third the size of the gross stock it is subtracted from, which gives it an awkward property. An error of one per cent in the liabilities alone, which for a stock built from surveys of millions of holdings is not demanding, moves the net position by some seven hundred billion dollars.²

That is not hypothetical. In June 2026 the Bureau of Economic Analysis published its annual update and restated the position for the end of 2025 from 27.54 trillion dollars to 21.87 trillion. Liabilities fell from 70.49 trillion to 64.78 trillion; assets were left nearly unchanged at 42.91 trillion.³⁴ The net figure shrank by 5.67 trillion dollars, about a fifth of the number it replaced, and no transaction caused any of it. What caused it was the arrival of the 2022 Benchmark Survey of Foreign Direct Investment in the United States, a 2023 insurance benchmark survey, the restatement of inward direct investment equity to market value for 2020 to 2025, and the discontinuation of an adjustment to loan positions.⁵

Benchmark surveys do this because of how the position is maintained between them. A census of foreign-owned businesses is occasional, so in between the stock is carried forward by cumulating recorded flows and applying valuation adjustments, which makes the published position for most years an extrapolation from the last count. It is also exposed to how direct investment is valued, because most countries report it at book value rather than market value, and valuation effects rather than transactions dominate short-run changes in net positions.⁶ A jump at a benchmark year is a feature of the method, not a failure of it. What is worth noticing is its size.

This would be a technical matter if nobody had built an argument on the number. For twenty years somebody has. The United States has been a net debtor on this measure since the late 1980s, yet for most of that period it earned more on its foreign assets than it paid on its liabilities, a persistence a recent Peterson Institute paper calls a longstanding puzzle.⁷ Ricardo Hausmann and Federico Sturzenegger proposed the boldest resolution: if the income flow says the United States earns like a creditor, perhaps it is one and the statistics are missing assets. They called the missing assets dark matter, after the astronomical inference from orbits to unseen mass, and attributed them to unmeasured intangibles and unrecorded exports of insurance and liquidity services, steady enough to keep the American net position stable and cast doubt on any need for a major dollar adjustment.⁸ The direction matters: a flow was used to re-estimate a stock, on the reasoning that the flow was measured better.

Pierre-Olivier Gourinchas and Helene Rey made a more careful version of the same observation, estimating an excess return on American foreign assets over liabilities of roughly two per cent a year in real terms and splitting it into a composition effect, from holding riskier assets against safer liabilities, and a return effect within each asset class.⁹ Critics went at the inference rather than the arithmetic. Stephanie Curcuru, Tomas Dvorak and Francis Warnock pointed out that the dark matter argument assumes the income data are more accurate than the service-export data, and that if so, service exports would be understated and the current account deficits overstated; they called that assumption's validity not entirely clear.¹⁰ A later criticism found the direct investment return gap came almost entirely from reinvested earnings rather than dividends.¹¹ Studies excluding other capital gains found much smaller differentials.⁹

Then the flow moved too. The same June 2026 update revised the 2025 primary income balance from an average quarterly surplus of 3.2 billion dollars to an average quarterly deficit of 11.9 billion.¹² The scarcity of that is the point: on one account the balance had been in deficit in only nine quarters since the series began in 1960.¹² The Federal Reserve Bank of New York, writing in May 2026, attributed the fading surplus to the rise in interest rates after the pandemic together with continued net sales of American assets to foreign investors, and put foreign holdings of American assets at 69 trillion dollars against American holdings abroad of 41 trillion.¹³ Those figures were assembled before the June revision cut the liabilities side by 5.71 trillion: a central bank's own account of the asymmetry rested on a number that moved the following month.

So the puzzle used to argue that the stock was mismeasured was itself argued from a flow, and that flow turns out to have been partly a measurement artefact and partly an interest rate environment. The income side also cannot stand in for the asset side. Gabriel Zucman estimated that about twenty per cent of American corporate profits were being booked in tax havens, and later work with Thomas Torslov and Ludvig Wier found that in 2015 American firms shifted more than half of their foreign profits.¹⁴ Haven-booked profits return to the accounts as American investment income, so the income balance is sensitive to where profits are declared rather than where anything was produced.

None of this makes the position useless. As a measure of gross financial integration it is excellent, and the gross stocks matter in their own right: a country with 47 trillion dollars of foreign assets and 69 trillion of liabilities is doing something different from one with a tenth of each and the same net figure. What it is not is precise in magnitude, and it carries a further peculiarity. American foreign assets are largely equity and largely in foreign currency, while American liabilities are largely dollar debt, so a fall in the dollar improves the position without a single transaction.⁹ A measure that moves when its own unit of account moves is doing something other than counting.

The strongest objection is that this amounts to nihilism about a number whose broad message has never wavered. The revision left the United States a net debtor of about 22 trillion dollars, as it has been for close to four decades, and the argument does not require otherwise. The sign is robust; the magnitude is not. That bites because the claims made with this number concern magnitude. Whether the position is sustainable, whether an exchange rate adjustment is needed, whether the reserve-currency privilege is worth two per cent a year or nothing, are questions about size, and a quantity whose level moves a fifth when a survey arrives cannot settle them.

Two habits follow. Report the gross positions alongside the net one, since they are measured more directly and the subtraction destroys information. And when a flow is used as evidence that a stock is mismeasured, ask what happens to the argument if the flow is revised. In June 2026 that question stopped being rhetorical.

Notes

1Bureau of Economic Analysis, "U.S. International Transactions and Investment Position, 2nd Quarter 2026," released 24 September 2026, for the net international investment position of -22.42 trillion dollars at the end of the second quarter of 2026, for assets of 46.97 trillion and liabilities of 69.39 trillion, and for the next scheduled release on 18 December 2026. The outbound network this paper was drafted on refused every connection to bea.gov, so the release was not opened at drafting and the figures came from search results quoting it. It was opened at review and carries them word for word: the position was -22.42 trillion dollars at the end of the second quarter of 2026, assets totalled 46.97 trillion and liabilities 69.39 trillion, and third-quarter statistics are scheduled for 18 December 2026. The release also gives the first quarter of 2026 as -21.27 trillion, revised. BEA news release and the release PDF

2This sentence is the author's own arithmetic on the figures in note 1: one per cent of 69.39 trillion dollars is 694 billion. It is an illustration of the sensitivity of a difference to its components, not an estimate of the actual error in the liabilities, which is not published. Underlying figures

3Bureau of Economic Analysis, "U.S. International Transactions and Investment Position, 1st Quarter 2026 and Annual Update," released 24 June 2026, for the revised net position of -21.87 trillion dollars at the end of the fourth quarter of 2025, for revised liabilities of 64.78 trillion against a previously published 70.49 trillion, for revised assets of 42.91 trillion, and for a first-quarter 2026 position of -21.27 trillion with assets of 43.37 trillion and liabilities of 64.64 trillion. The 5.67 trillion dollar difference between -27.54 and -21.87 trillion, and the 5.71 trillion reduction in liabilities, are the author's subtractions from these published figures. Not opened directly; figures quoted from search results returning this release and its PDF. BEA news release and the release PDF

4Bureau of Economic Analysis, "U.S. International Transactions and Investment Position, 4th Quarter and Year 2025," for the pre-revision estimates of the position, reported as -27.61 trillion dollars preliminary for the fourth quarter of 2025 and -27.55 trillion revised for the third quarter, against -26.54 trillion at the end of 2024. The baseline for the revision therefore depends slightly on which table is used, which is why the comparison in the text uses the -27.54 trillion figure carried in the June 2026 release itself. BEA news release and the St Louis Fed series, which carries the revised levels

5Bureau of Economic Analysis, "2026 Annual Update of the U.S. International Economic Accounts," Survey of Current Business, July 2026, for the incorporation of the 2022 Benchmark Survey of Foreign Direct Investment in the United States and a 2023 insurance benchmark survey, for the revision of direct investment positions for 2022 to 2025, for the restatement of foreign direct investment equity in the United States to market value for 2020 to 2025, for the discontinuation from 2023 of a direct investment-related adjustment to other investment loan positions, and for new statistics on stock swaps from 2023. Not opened; the apps.bea.gov host was unreachable from this network. Survey of Current Business article and BEA's Concepts and Methods document, September 2026

6Philip R. Lane and Gian Maria Milesi-Ferretti, "The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970-2004," IMF Working Paper WP/06/69, for the construction of positions from stock measures supplemented by cumulated flows with valuation adjustments, for the observation that the majority of countries report direct investment at book value with only a small number reporting market values, and for the quantitative importance of valuation effects. The point that valuation effects dominate short-run changes in net positions while tending to offset over long horizons, and that the 2024 widening of creditor and debtor positions was driven primarily by equity price changes, is taken from the Brookings update. IMF working paper, a hosted copy of the published version and the External Wealth of Nations update to year-end 2024

7Peterson Institute for International Economics working paper 25-14, "The US Trade Deficit and Foreign Borrowing," for the description of the persistence of positive net investment income despite a negative net international investment position as a longstanding puzzle in the US external accounts. Working paper

8Ricardo Hausmann and Federico Sturzenegger, "The Missing Dark Matter in the Wealth of Nations and Its Implications for Global Imbalances," Economic Policy 22, no. 51 (July 2007), and the earlier Harvard Center for International Development working paper version, for the dark matter concept as the gap between the net foreign assets implied by American investment income and the official figure, for the attribution to unmeasured intangibles and unreported exports of insurance and liquidity services, and for the conclusion that steady dark matter exports kept the net position stable and cast doubt on the need for a major dollar adjustment. Harvard Kennedy School working paper copy, the author's own copy of the Economic Policy article and a bibliographic record

9Pierre-Olivier Gourinchas and Helene Rey, "U.S. External Adjustment and the Exorbitant Privilege," for the excess return on American foreign assets over liabilities of around two per cent a year in real terms, for the decomposition into a composition effect and a within-class return effect, for the role of valuation adjustment in external adjustment given that American assets are weighted towards foreign-currency equity while liabilities are weighted towards dollar debt, and for the existence of studies that, excluding other capital gains, find significantly lower differentials. The companion argument that the United States also bears an exorbitant duty, taking valuation losses in crises, is in the Gourinchas, Rey and Govillot paper. NBER chapter, Rey's own copy of the exorbitant duty paper and Bertaut, Curcuru, Faia and Gourinchas, new evidence on the US excess return, Federal Reserve IFDP 1398

10Stephanie E. Curcuru, Tomas Dvorak and Francis E. Warnock, "Cross-Border Returns Differentials," Board of Governors of the Federal Reserve System, International Finance Discussion Paper 921 (2008), for the argument that the dark matter inference depends on treating the balance-of-payments income data as more accurate than the service export data, for the implication that service exports would then be understated and American current account deficits overstated, and for the authors' judgement that the validity of this assumption is not entirely clear. Not opened; the characterisation and the quoted phrase were returned in search results quoting the paper's own text. Federal Reserve paper

11Council on Foreign Relations, "Searching for the origins of dark matter," for the finding that the gap in direct investment returns comes almost entirely from reinvested earnings rather than dividends and that the dark matter assumptions about insurance income do not match the data, and the companion piece for the observation that anything reducing the measured size of the net position mechanically reduces the implied stock of dark matter. Searching for the origins of dark matter and Dark matter watch

12Haver Analytics, "Q1 Current Account: Modest Slippage in Early 2026," for the revised 2025 primary income balance showing an average quarterly deficit of 11.9 billion dollars against a previously published average quarterly surplus of 3.2 billion, and the same firm's earlier note for the statement that the primary income balance had been in deficit in only nine quarters since the series began in 1960. Note that BEA's own revised figure for the fourth quarter of 2025 is a primary income surplus of 3.4 billion dollars, which is consistent with an annual average deficit only if other quarters were in deficit; the quarterly and annual figures are reported here as they stand rather than reconciled. Haver on the Q1 2026 account and Haver on the Q2 2025 account

13Matthew Higgins and Thomas Klitgaard, "Honey, Who Shrunk the U.S. Income Surplus?," Liberty Street Economics, Federal Reserve Bank of New York, 18 May 2026, for the attribution of the fading net investment income surplus to the upward shift in interest rates after the pandemic together with continued net sales of American assets to foreign investors, and for the figures of 69 trillion dollars of foreign holdings of American financial assets against 41 trillion dollars of American holdings of foreign assets. Liberty Street Economics post and a repository record

14Gabriel Zucman, "Taxing across Borders: Tracking Personal Wealth and Corporate Profits," Journal of Economic Perspectives 28, no. 4 (Fall 2014), pp. 121-148, for the estimate that about twenty per cent of all American corporate profits were being booked in tax havens, a tenfold increase since the 1980s; and Thomas Torslov, Ludvig Wier and Gabriel Zucman, "The Missing Profits of Nations," Review of Economic Studies (2022), for the estimate that 36 per cent of multinational profits are shifted to tax havens globally and that in 2015 American firms shifted more than half of their multinational profits. The inference drawn in the text, that haven-booked profits returning as American investment income make the income balance sensitive to where profits are declared, is the author's rather than a finding stated in either paper. Zucman's own copy of the JEP article and his copy of the Review of Economic Studies paper