# Relative Wealth Signals for the "Renegade" G20 Economies: A Methodological Investigation

*Working brief — StratMac Research Library*

**Date:** 2026-08-07
**Status:** Research memo (peer-review-ready methodology documentation); companion to the site's data-correctness policy.
**Principal author:** DeepSeek v4 Flash — authored this investigation, the build steps, and the empirical validation recorded here.
**Advisory:** Mike Landis — directed scope, mandated honesty-over-publication, and reviewed conclusions.
**Contact:** StratMac Research Library — Strategic.Macroeconomics@gmail.com
**Classification:** Open publication. This brief exists precisely so the methodology — including its dead ends — is a matter of public record, not a private assumption.

---

## 1. Purpose and scope

StratMac publishes a *Global Wealth per Capita* card for each G20+ economy: the country's **wealth-per-adult divided by U.S. GDP-per-capita**. Three economies have refused to yield an honest value of that quantity from any source we can verify — **Turkey, Argentina, and the European Union aggregate** (which shares the same data dependency). Their cards were **unpublished in 2026-08** after the per-adult series they were built on was found to carry a ~10³–10⁴× unit-scale error that produced impossible, reader-visible spikes ("crashes" of 39×, 1005×, 22.9× vs. U.S. GDP-per-cap).

This brief documents, for the record, **every source we evaluated and every USD-conversion method we attempted**, the empirical validation we ran against trusted anchors, the structural reason each failed, and why we chose to leave these cards unpublished rather than ship a synthetic series.

---

## 2. The target quantity and why it compounds a known error

The published quantity is

$$\text{WEALTCAP}_c(t) = \frac{W_c(t)} {\text{USGDPcap}(t)}$$

where $W_c(t)$ is country $c$'s mean household **wealth per adult** in current U.S. dollars at year $t$, and $\text{USGDPcap}(t)$ is U.S. GDP per capita. The numerator crosses currencies for every non-U.S. country, so the entire signal depends on a defensible **local-currency → USD conversion** of a multi-decade wealth series.

The 10³–10⁴× error in Turkey/Argentina originated exactly here: a local-currency wealth series was divided by a currency-incompatible dollar rate, inflating (or collapsing) the ratio by orders of magnitude. Rebuilding the card therefore required, first, a source with trustworthy per-year USD wealth-per-adult; we found none that is both continuous **and** USD **and** verifiable for these economies — and the reasons are specific and informative.

---

## 3. Sources evaluated (complete list)

| # | Source | Continuous (per-year)? | USD? | Verifiable openly? | Coverage | Verdict |
|---|--------|----------------------|------|-------------------|----------|---------|
| A | **UBS / Credit Suisse Global Wealth Databook** | ✗ (snapshot years end-2000, end-2023) | ✓ | ✗ PDFs bot-blocked (403/HTML shell) | global | authoritative USD but not a time series |
| B | **WID (World Inequality Database, wid.world)** | ✓ (annual, ~1800–2025) | ✗ (local currency / PPP) | ✓ (bulk CSV now staged) | global | continuous but currency-native |
| C | **Wikipedia "List of countries by wealth per adult"** (collates UBS) | ✗ (single-year end-2022) | ✓ | ✓ (HTML table) | global | USD but one year only |
| D | **datahub.io rehost** | ✗ | ✓ | partial (Vercel 404 on raw CSV) | 20 countries | insufficient |
| E | **IMF WEO / World Bank ICP (PPP)** | ✓ (annual) | ~ (PPP units) | ✓ | global | PPP, not market USD |

**Key structural fact:** WID is the only source that gives *continuous* per-year wealth-per-adult, and it does so only in **local currency** (or WID's own PPP construct). Every other continuous source is PPP or not USD; every USD source is a snapshot. No single verifiable source provides continuous USD. The gap must be bridged.

---

## 4. Methods attempted, with empirical validation

We staged WID's full bulk dataset (`wid_all_data.zip`, 882 MB; landed in `data_source/wid/`, dated 2026-07-25 release) and worked from the raw per-country CSVs. The correct variable is `ahweali992` = **wealth per adult, full population** (the `999` suffix is per-*capita* — a trap we confirmed and rejected). We validated every conversion against trusted **UBS 2022 mean wealth-per-adult anchors** (USD): US 551,347 · DE 256,179 · CH 685,226 · CN 75,000 · TR 17,578 · AR 13,323 · BR 16,700 · RU 26,896.

### 4.1 Attempt 1 — Single-year market FX
Divide WID local-currency wealth by a current-year market FX rate. **Result:** reproduces stable-currency economies (US 0.98×, CN 1.03×, DE 0.85×) but over-shoots high-inflation economies wildly (TR 6.1×, AR 41× at 2022). **Cause:** the wealth series is a real/deflated LCU construct; a nominal current FX is unit-mismatched.

### 4.2 Attempt 2 — Single-year WID PPP converter (`xlcuspi999`)
Divide by WID's own LCU-per-PPP-USD. **Result:** similar failure (TR 20×, AR 89× at 2022). Same root cause: PPP converter is nominal-base, not matched to the real wealth series.

### 4.3 Attempt 3 — Per-year market FX (`xlcusxi999`) and per-year PPP (`xlcuspi999`)
Use each year's own converter, paired with each year's wealth value. **Result:** still mis-stated for high-inflation countries — TR 6–20×, AR 41–89×, BR 2.6×, RU 2.8× off target; stable-currency CN/US/DE/CH/FR land within 0.85–1.9×. The FX history is fully available; it simply does not reconcile with the wealth base for these economies.

### 4.4 The structural diagnosis (why 4.1–4.3 fail)
For high-inflation economies, WID's `ahweali992` is a **real (price-deflated) local-currency series**. Evidence — Turkey: nominal wealth moves only 945k → 1.8M (1990→2023), while the TRY/USD FX explodes 0.003 → 39 over the same window. A real-value numerator divided by a nominal-exchange denominator is **dimensionally inconsistent**. The handful of economies whose nominal FX closely tracks their real price level (US, DE, CH, CN, FR) happen to convert cleanly; the high-inflation ones cannot. This is precisely the error family that corrupted the original Turkish/Argentine files at ~10³–10⁴×.

### 4.5 Attempt 4 — FX-free within-currency ratio (not deployed)
Form $\frac{W_c(t)}{GDP_c(t)}$ — country wealth-per-adult ÷ **own** GDP-per-adult, both real LCU, both from WID. This is the only conversion that is **provably currency-consistent** for every country, because the currency cancels identically in numerator and denominator. It is statistically sound and corruption-proof by construction. **Not adopted** because it redefines the card from "wealth vs *US* GDP" to "wealth vs *own* GDP" — a change of semantics we regard as a reader-facing methodological decision, not one an implementation should make unilaterally.

---

## 5. Findings (encapsulated)

1. **No verifiable, openly-accessible, continuous-per-year USD wealth-per-adult source exists** for Turkey, Argentina, or the EU aggregate. This is a genuine gap in the public statistical infrastructure, not a gap in StratMac's tooling.
2. **WID's wealth-per-adult is real-value local currency**, and for high-inflation economies its conversion to USD via *any* nominal FX or PPP rate is **dimensionally invalid**. We reproduced the original ~10⁴× corruption mechanism in controlled attempts.
3. **Stable-currency economies convert cleanly** from WID (validated 0.85–1.9× against UBS), which is why 35 of 38 global-wealth cards are sound and stay published.
4. **There is a provably-correct alternative** — the within-currency wealth/GDP ratio — but it changes the card's meaning and is therefore held for an explicit methodological decision rather than silently substituted.
5. **Publishing nothing is the correct near-term state** for these three cards. It is preferable to a number that is wrong by one or two orders of magnitude, and the site's integrity rests on readers being able to trust the cards we *do* show.

---

## 6. Decision and disposition

- **Turkey (`G20_TUR`), Argentina (`G20_ARG`):** are **unpublished**. Their prior data is withdrawn. Restoration is gated on a verifiable currency bridge (the within-currency ratio being the only mechanism shown to be structurally sound) or a new open USD source.
- **EU aggregate (`G20_EUU`):** is **unpublished**; it carries the same data dependency and will be restored by the same path.
- **Brazil / Russia / Romania:** their *visible* (post-2000) windows are plausible (2.4×, 4.8×, 2.2×), so their cards stay up. Their **pre-2000** values carry the same unit-scale corruption history, which is invisible to readers; this is logged and flagged for correction by any future regenerated pipeline, but is not a reader-facing defect today.
- The `sm-card-corruption-watchdog` (daily, script-only) remains active and will alert if any visible-window value ever exceeds 15× or falls below 0.02× of U.S. GDP-per-cap, or spikes/crashes steeply year-over-year.

---

## 7. Reopening criteria and open questions

Those who wish to revisit this should know what would change our conclusion:

- **A new continuous USD source** (official-currency WID variant, or a documented real-FX deflation of the WID wealth series validated against the UBS anchors at both end-2000 and end-2023) would permit faithful restoration of the original card semantics.
- **An explicit decision to adopt the within-currency ratio** (wealth ÷ own GDP) would let all three return immediately with currency-consistent data.
- **Open methodological question for the advisory:** Is "wealth ÷ own GDP" a *different, acceptable* signal, or is only "wealth ÷ US GDP-per-cap" acceptable? This is the crux; the data cannot decide it, only a policy decision can.

---

## 8. Sources of primary data cited

- WID, *World Inequality Database*, full bulk export (2026-07-25), var `ahweali992`, `xlcusxi999`, `xlcuspi999`.
- UBS / Credit Suisse, *Global Wealth Report / Databook* (2022, 2023, 2024 editions), mean wealth per adult (USD).
- Wikipedia, *List of countries by wealth per adult* (UBS-collated table, end-2022) — used as the anchor table.

*This brief is part of the StratMac data-correctness policy record. Full policy: `multi-site-web-ops/references/data-correctness-priority.md`; catalogue: `papers/catalogue.md`.*