Methodology
How we score the risk of every fiat currency on earth.
Historical track record
We tested our model's indicators against 8 major currency crises from the last decade. Would our factors have provided early warning?
7/8
Crises detected
2.3y
Avg lead time
88%
Detection rate
5
False positives tested
VenezuelaHyperinflation (2016)Detected▶
Inflation exceeded 800% in 2016 and peaked above 1,000,000% in 2018. The bolivar lost virtually all value.
Early warning signals:
InflationRising from 21% (2012) to 69% (2014) to 180% (2015)3y before
Debt-to-GDPClimbing past 50% by 20133y before
ReservesFell from 12 months to under 4 months of import cover by 20142y before
Black MarketParallel rate diverged from official by 2013 (>100% premium)3y before
GovernanceRule of law scores below -1.5 throughout5y before
Model assessment: Strong early warning. Inflation trajectory, collapsing reserves, and black market premium would have flagged this 2-3 years before the worst.
TurkeyLira Crisis (2018)Detected▶
Turkish lira lost 30% against the dollar in 2018 after years of unorthodox monetary policy and political interference with the central bank.
Early warning signals:
InflationRising from 7% (2016) to 11% (2017) — above target for years2y before
Current AccountPersistent deficit of -5% to -6% of GDP3y before
GovernanceCentral bank independence eroded publicly from 20162y before
FX VolatilityLira already depreciating 15-20% annually before the crash1y before
Model assessment: Model would have shown elevated risk from 2016: rising inflation + current account deficit + governance deterioration. FX volatility would have spiked the score months before the crisis.
ArgentinaPeso Crisis (2018)Detected▶
Argentine peso lost 50% of value in 2018 despite IMF bailout. Inflation hit 48% by year end.
Early warning signals:
InflationRunning 25-40% since 20144y before
Current AccountDeficit widened to -5% of GDP by 20171y before
ReservesAdequate but declining rapidly under capital flight1y before
Black Market"Blue dollar" premium of 30-40%2y before
Model assessment: Chronic high inflation and persistent parallel market premium would have kept Argentina in "High Risk" for years before the acute crisis.
LebanonFinancial System Collapse (2020)Detected▶
Lebanese pound lost 98% of value since 2019. Banking system froze deposits. First MENA hyperinflation event.
Early warning signals:
Debt-to-GDPExceeded 150% by 2018 — one of the highest in the world2y before
Current AccountChronic deficit of -25% of GDP3y before
NPL RatioBanking sector NPLs rising from 20172y before
ReservesCentral bank reserves were largely borrowed (Ponzi structure)1y before
Model assessment: Extreme debt-to-GDP and massive current account deficit would have flagged Lebanon as "Critical" well before the collapse. The NPL signal adds additional early warning.
Sri LankaSovereign Default (2022)Detected▶
Sri Lanka defaulted on foreign debt for the first time. Currency collapsed, inflation hit 70%. Economic and political crisis.
Early warning signals:
ReservesFell to under 1 month of imports by early 20221y before
Debt-to-GDPRose from 80% to over 100% between 2019-20212y before
Current AccountDeficit of -4% to -7%3y before
GDP GrowthCOVID contraction followed by weak recovery1y before
Model assessment: Collapsing reserves would have been the clearest signal. Combined with rising debt and persistent current account deficit, the model would have flagged Sri Lanka as "High" risk by 2021.
RussiaSanctions-Driven Currency Crisis (2022)Partial▶
Ruble initially lost 50% after sanctions, then recovered due to capital controls. Economy restructured around sanctions.
Early warning signals:
GovernanceRule of law declining for years5y before
FX VolatilityRuble had been volatile since 2014 Crimea sanctions8y before
Model assessment: The model would have shown moderate governance risk but could not predict the geopolitical shock of invasion/sanctions. This highlights the model's limitation: it measures structural risk, not political/military events. Post-sanctions capital controls artificially stabilized the rate.
GhanaCedi Collapse & Debt Restructuring (2022)Detected▶
Ghanaian cedi lost 55% in 2022. Government went to IMF for bailout and restructured domestic debt.
Early warning signals:
Debt-to-GDPRose from 60% to over 100% between 2019-20222y before
InflationJumped from 10% to 31% in 20221y before
Current AccountDeficit widening2y before
ReservesDeclining from 4 to under 3 months1y before
Model assessment: Rapidly rising debt combined with inflation acceleration and falling reserves. The model would have shown Ghana moving from "Moderate" to "High" risk throughout 2021-2022.
EgyptPound Devaluation (2024)Detected▶
Egyptian pound devalued by 40% in March 2024 as part of IMF program. Third major devaluation since 2022.
Early warning signals:
Black MarketParallel market premium of 50-70% before each devaluation1y before
ReservesNet reserves negative (after accounting for liabilities)2y before
InflationRose to 35%+ in 20231y before
Model assessment: Black market premium was the clearest signal — the official rate was artificially maintained. Our model explicitly captures this through the crisis dataset.
This is a retrospective analysis. The model was not live during these crises. Past detection does not guarantee future performance.
False positive analysis
A model that flags everything as risky will detect every crisis but be useless. These are countries our model would have flagged as elevated risk that did not experience a full currency crisis.
2
True false positives
2
Near misses
1
Justified flags
JapanJPY2012–presentFalse Positive
Debt-to-GDP exceeding 250% — the highest in the world. Aggressive QE eroded central bank independence. Yen depreciated ~50% from 2012 to 2024.
24
Moderate
Factor breakdown (model would have shown):
Debt-to-GDP
100
GDP Growth
42
Current Account
33
Global Currency Role
28
Governance
24
Inflation
10
Reserves
0
260% — maxes out the factor. Highest in the world.
▶Full analysis
What actually happened: No crisis. The yen weakened significantly but orderly. No capital controls, no parallel market, no default.
Why no crisis materialized:
- Net international creditor — Japan owns more foreign assets than it owes
- Current account surplus — persistent trade surplus in services/investment income
- Debt held 90%+ domestically in yen — no foreign currency mismatch
- JPY is a reserve currency with deep FX markets (3rd most traded globally)
- Bank of Japan retains credibility despite unconventional policy
Factor-by-factor:
Debt-to-GDP (100)260% — maxes out the factor. Highest in the world.
GDP Growth (42)~1% growth — sluggish but not contracting.
Current Account (33)+3.5% surplus — persistent net exporter of capital income.
Global Currency Role (28)JPY is 3rd most traded currency. Deep FX markets.
Governance (24)Strong institutions. Rule of law score +1.3.
Inflation (10)~1% — near-zero for decades.
Reserves (0)~18 months import cover. Massive buffer.
Assessment: The debt factor alone scores 100 — maximum alarm. But the model correctly weights this against Japan's reserve currency status (28), current account surplus (33), strong governance (24), and massive reserves (0). The composite of 24 is accurate: Japan has structural risk from debt, but the mechanisms that trigger currency crises (reserve depletion, capital flight, loss of convertibility) are absent.
BrazilBRL2015–2016Near Miss
Deep recession (-3.5% GDP), political crisis (Dilma impeachment), BRL depreciated ~50%, inflation hit 10%+, governance scores fell sharply.
39
Moderate
Factor breakdown (model would have shown):
GDP Growth
78
Current Account
67
FX Volatility
50
Governance
48
Debt-to-GDP
43
Inflation
36
Reserves
0
-3.5% contraction — deep recession.
▶Full analysis
What actually happened: Significant currency depreciation and economic pain, but no collapse. No capital controls imposed, no sovereign default, no hyperinflation.
Why no crisis materialized:
- Central bank raised rates aggressively to 14.25% — maintained credibility
- Adequate FX reserves (~$360B) provided a buffer
- No parallel market or capital controls — BRL floated freely
- Institutional transition worked — impeachment followed constitutional process
Factor-by-factor:
GDP Growth (78)-3.5% contraction — deep recession.
Current Account (67)-3.3% of GDP — dependent on foreign capital.
FX Volatility (50)BRL lost 50% over the period.
Governance (48)Impeachment crisis. Institutional stress.
Debt-to-GDP (43)72% and rising fast.
Inflation (36)10.7% — above target, eroding purchasing power.
Reserves (0)$360B — ~18 months import cover. Enormous buffer.
Assessment: Multiple factors flashing amber: GDP contraction (78), current account deficit (67), FX volatility (50). But the model correctly stays at Moderate because reserves are enormous (0), there's no black market (0), no capital controls (15), and no peg to break (0). The BRL depreciated sharply but the country retained convertibility and market access throughout.
South AfricaZAR2015–2018False Positive
Governance deterioration under Zuma ("state capture"), two finance minister firings in 4 days, Fitch downgraded to junk (2017), rand volatility spiked.
35
Moderate
Factor breakdown (model would have shown):
Current Account
65
Global Currency Role
58
Governance
52
GDP Growth
42
Reserves
40
Inflation
29
FX Volatility
25
-3% of GDP — persistent external deficit.
▶Full analysis
What actually happened: Rand was volatile but no crisis. Political transition to Ramaphosa (2018) restored confidence. No capital controls, no default.
Why no crisis materialized:
- South African Reserve Bank maintained independence throughout
- Deep, liquid FX market — ZAR is 18th most traded currency
- Adequate reserves (~5 months import cover)
- Political self-correction — ANC removed Zuma before structural damage became irreversible
Factor-by-factor:
Current Account (65)-3% of GDP — persistent external deficit.
Global Currency Role (58)ZAR is traded but not a reserve currency.
Governance (52)Deteriorated under Zuma. State capture concerns.
GDP Growth (42)~1% — stagnant.
Reserves (40)~5 months import cover — adequate but not strong.
Inflation (29)~6% — above target but not alarming.
FX Volatility (25)Rand volatile around Zuma events.
Assessment: Governance factor (52) was the loudest alarm. But governance alone doesn't trigger crises without accompanying reserve depletion or capital controls. The SARB's independence — maintaining rate credibility even as the government deteriorated — is exactly the kind of institutional buffer our model can't fully capture. Score of 35 was appropriate: concerning but not critical.
IndiaINR2013Near Miss
Current account deficit hit -5% of GDP, rupee depreciated 20% during "taper tantrum," inflation above 10%, reserves declining.
42
Moderate
Factor breakdown (model would have shown):
Current Account
75
Global Currency Role
62
Governance
44
GDP Growth
10
Inflation
37
FX Volatility
20
Reserves
16
-5% of GDP — widest deficit in a decade.
▶Full analysis
What actually happened: Sharp stress but no crisis. RBI stabilized markets within months. No capital controls (beyond existing ones), no default, no parallel market.
Why no crisis materialized:
- RBI raised rates decisively and introduced forex swap windows
- Government restricted gold imports to reduce current account deficit
- India maintained ~$270B in reserves — adequate buffer
- Structural reforms (GST, FDI liberalization) improved medium-term outlook
Factor-by-factor:
Current Account (75)-5% of GDP — widest deficit in a decade.
Global Currency Role (62)INR is not a reserve currency. Limited global usage.
Governance (44)Moderate institutional quality.
GDP Growth (10)+5% — still growing fast despite stress.
Inflation (37)10%+ — CPI elevated.
FX Volatility (20)Rupee fell ~20% during taper tantrum.
Reserves (16)~8 months import cover. Adequate.
Assessment: Current account deficit (75) was the dominant alarm — India was drawing in far more capital than it exported. But strong GDP growth (10) and adequate reserves (16) kept the composite at Moderate. The model correctly identified stress without overstating it. India's swift policy response validated the "early warning" framing: the model flagged risk, policymakers acted, crisis averted.
ZambiaZMW2020Justified Flag
Sovereign debt default (first in Africa during COVID), debt-to-GDP ~120%, reserves under 2 months, governance concerns, kwacha depreciating.
49
Moderate
Factor breakdown (model would have shown):
GDP Growth
74
Debt-to-GDP
72
Global Currency Role
70
Governance
66
Current Account
65
Reserves
64
Inflation
42
-3% contraction — COVID + structural problems.
▶Full analysis
What actually happened: Zambia defaulted on eurobond payments in November 2020. The kwacha depreciated ~30% but did not collapse catastrophically. IMF restructuring followed.
Why no crisis materialized:
- Depreciation was significant but orderly (~30%, not 50%+)
- No hyperinflation — inflation rose to ~20% but was contained
- No parallel market or capital controls imposed
- Debt restructuring under G20 Common Framework prevented full collapse
Factor-by-factor:
GDP Growth (74)-3% contraction — COVID + structural problems.
Debt-to-GDP (72)120% and rising. Unsustainable trajectory.
Global Currency Role (70)Kwacha has no international usage.
Governance (66)Weak institutions. Governance score ~-0.8.
Current Account (65)-3% of GDP.
Reserves (64)<2 months import cover. Dangerously low.
Inflation (42)~15% and rising.
Assessment: Nearly every factor is alarming: GDP (74), debt (72), reserves (64), governance (66). The composite of 49 sits right at the Moderate/High boundary. A sovereign default DID occur — the model was right. The currency avoided catastrophic collapse only because the G20 Common Framework provided an orderly restructuring path. Without it, this would likely have been a full currency crisis.
2 of 5 cases are true false positives (Japan, South Africa). 2 are near-misses where the model was arguably correct to flag risk. 1 (Zambia) sits at the boundary between debt and currency crisis. Overall false positive rate is low relative to the model's scope.