How this site is built and where every number comes from

A thin third app (shape A in the plan of 31 October 2026, refurbished on 2 November 2026). The funding side is money-flow’s export surface: the Signal Board (method 1.4), the economy views, the earnings series (World Bank WDI and IMF quarterly, fetched by money-flow), its change feed and its validation set, plus the links tables from BIS and IMF statistics. The trade side is trade-flow’s export surface: UN Comtrade 12-month windows, partners, chain stages, HS-chapter products, its change feed and the trade events. Nothing is fetched from a publisher here, so nothing can disagree with the two source sites; what a parent shows in full is linked, not copied.

The economy set is money-flow’s board rule: the 36 borrowing economies of the Signal Board, Korea first. The United States is the hub of the dollar system, not a borrower in this sense; it appears for trade and on the pulse only, and has no score.

Nineteen rules decide what may be joined and shown. G11–G17 were the Fiscal Stress Monitor’s and moved with it to money-flow (rules R22–R28 there); G18 is the boundary rule of the refurbish of 2 November 2026; G19 is the stress signals’ admission rule of 3 November 2026. Each page cites the rule it follows.

The Overview splits each economy’s current account into its two earning lines. Across: the balance on goods and services, what it earns by selling abroad less what it pays for what it buys. Up: the primary income balance, the interest, dividends and reinvested profits it earns on its foreign assets less what it pays foreigners on their assets inside it (wages of cross-border workers are in it too, and small). Both are divided by GDP in current dollars, a year at a time. Secondary income, mostly remittances and aid, is the third line of the current account and is left out.

Quadrants: a balance above 1% of GDP counts as earning on that axis, one between −1% and 1% as not earning. So an economy “earns by trading” with a trade surplus above 1% and primary income at or below 1%, “earns from foreign assets” the other way round, “earns both ways” with both above, and “pays out both ways” otherwise. The band keeps a surplus of a few tenths of a percent from counting as a way of living; it is a reading aid, not a model.

Bubble size is GDP per capita in current dollars; colour is the net international investment position at the last published quarter of the year (money-flow, IMF IIP), in five bins. These numbers are shown and never scored: they do not enter V or F, and nothing here recomputes either. The side panel joins them with trade-flow’s export market concentration and world share (12 months to the shown month), the chokepoints the economy holds (the last complete year) and A1 (the latest quarter), each at its own period.

Two more views show the scale the net balances hide. Trade scale plots exportsⓘ against importsⓘ of goods and services, both on the balance-of-payments basis of the net line and divided by the same World Bank GDP, so a point’s distance straight down to the 45° lineⓘ is exactly its trade balance in the Net view. Financial scale plots gross external assetsⓘ against gross external liabilitiesⓘ from the IMF IIP at the year’s last published quarter, on a log scaleⓘ by default because financial centres reach five to six times GDP. Bubbles can be sized by GDP per capita, exports in dollars or gross assets in dollars; the side panel ranks each economy’s exports and gross assets among the economies shown. None of this enters V or F.

Why the World Bank and not the IMF files money-flow mirrors: the IMF balance-of-payments files in money-flow’s hub carry these lines, but their currency labels do not reconcile for several economies (Germany’s “euro” primary income equals the dollar figure). The World Bank republishes the IMF’s series in current dollars with GDP beside them; Japan’s primary income matches the IMF dollar series exactly. Since the refurbish of 2 November 2026 money-flow fetches both the WDI and the IMF quarterly series and publishes them in its earn/global view; the annual-against-quarterly GDP check is money-flow’s V23, read back on this page.

AUROC is the chance that a randomly drawn pre-episode quarter scores higher than a randomly drawn calm quarter. 0.5 is a coin toss; 1.0 would put every stressed quarter above every calm one. Early-warning scores in the literature sit around 0.6 to 0.8.

Intervals: 1000 bootstrap redraws of the economy-episodes (the positives) against the fixed calm set; the central 90%. A gap between two indicators that is smaller than their intervals is not a gain.

The honest reading: V and inverted export momentum, equal weights is highest in this build (0.79 (0.73–0.84)), but on 17 positives. The trade indicators are percentiles of an economy’s own history that needs eight quarters, and the 12-month windows begin in 2006, so the 2008 episodes drop out of their test. Its interval overlaps V’s, so the gain over V alone is not established; the trade indicators stay beside the score and not in it (G6).

Why V reads 0.725 here and 0.686 in money-flow’s method 1.4 memo: the test differs in two places. The memo’s calm set is every quarter more than four quarters from an economy’s episodes; ours leaves out ±8 quarters, so fewer near-episode quarters count as calm. The memo scores each of the four pre-episode quarters as a positive; ours takes the highest of the four, one positive per episode, which asks “did the score rise at some point in the year before” rather than “was it high every quarter”. Both choices raise the figure. The memo’s number is the reference for method 1.4; this one is for comparing indicators on the same footing.

• Coverage: 36 economies, the Signal Board’s set. Monthly trade is partly estimated where importers have not filed, and marked where it is.
• Derived flows carry valuation effects: a flow taken as the change of a position moves with exchange rates and prices as well as with transactions. BIS adjusted changes are used where published.
• B1 includes interbank funding, so for financial-centre advanced economies (the UK, Japan, Canada) the flow score reads bank-to-bank positions as withdrawals.
• A2 is shown but not scored for the UK, Japan, Canada and euro members: reserve issuers and economies with standing dollar swap lines do not fund short-term bank debt from reserves.
• The BIS debt securities currency split is sparse for many economies, so no bond figure is shown by currency.
• The AUROC intervals redraw the economy-episodes only; the calm set is fixed and the episodes were chosen by hand, so the intervals understate the uncertainty of the list itself. A mechanical crash dating gave money-flow a lower figure (0.59).

Every build records the money-flow release and the trade-flow build it read, and the version of each export surface. These are the figures of this build.

Cite the site, the page and the build date. The build date is the day this site joined the two surfaces; the money-flow release and trade-flow build it read are in the footer of every page and above.

datasteg global-network, “<page title>”, built <build date>, <page URL>.

datasteg global-network, “About”, built 2 Nov 2026, https://global-network.datasteg.com/en/about.

For the underlying figures cite the publisher and, where a number was derived, the parent: money-flow release and method version for funding, trade-flow build for trade.