IESET.
Hypotheses·regulatory·banking_crisis_japan_1990_lost_decade

Japan's 1990-2003 banking-and-asset-bubble bust — Nikkei index decline of >= 60% peak-to-trough, real residential property price decline of >= 40%, persistent bank-NPL accumulation that required government recapitalisation in 1998 and 2003, and a "lost decade" of GDP growth averaging <= 1% — is a canonical case of a delayed-resolution banking crisis.

The hypothesis is that Japan 1990-2003 meets the canonical multi-metric checklist on at least 4 of 5 metrics.

SUPPORTEDengine/runs/banking_crisis_japan_1990_lost_decade

SUPPORTED

confidence cueThis is a clear pass for the claim as written. It still applies only to this sample, period, and method.

policy briefClear support

In ordinary language

In plain terms, this asks whether the policy story survives a real-world data check from 1985 to 2005.

plain answer

The data clearly moved in the predicted direction. SUPPORTED

why it matters

This matters because regulatory claims should change belief only when they survive a pre-declared empirical test.

how the test works

It compares 1 country or place units from 1985 to 2005, using a multi metric checklist design.

what was measured
What we checked
  • Real income
  • Real house price index
what this does not prove

A single test is not the whole truth. It narrows the claim under a specific sample, time period, and method. Strong policy conclusions need the pattern to survive nearby tests, alternative data, and serious objections.

verification

5 input datasets, 0 unresolved missing series, provenance status: reproducible hash verified.

Results

engine/runs/banking_crisis_japan_1990_lost_decade
1007550250198519952005JPN
illustrative sketch · run pending
No coefficients yet. When the model fires, this chart will show real_gdp across 1 sampled countries over 1985–2005.
The shapes above are stylised — none of the lines are real data.
Placeholder for banking_crisis_japan_1990_lost_decade. Published chart will be generated from engine/runs/banking_crisis_japan_1990_lost_decade/chart_data.json.

Pre-registration

pre-registered
first-spec commit 098ce96 · 2026-04-30T12:57:33Z
run generated · 2026-05-15T20:30:32Z

Japan's 1990-2003 banking-and-asset-bubble bust — Nikkei index decline of >= 60% peak-to-trough, real residential property price decline of >= 40%, persistent bank-NPL accumulation that required government recapitalisation in 1998 and 2003, and a "lost decade" of GDP growth averaging <= 1% — is a canonical case of a delayed-resolution banking crisis. The hypothesis is that Japan 1990-2003 meets the canonical multi-metric checklist on at least 4 of 5 metrics.

Falsification criterion — what would disprove this

set before the run · honoured after

This hypothesis is considered falsified if:

Evaluate every canonical_metrics row against its pre-registered source, window, and threshold. The hypothesis is SUPPORTED if at least 4 of 5 metrics are MET. It is REFUTED if even counting all pending metrics as favorable cannot reach 4 MET metrics and the confirmed failures cross the pre-registered refutation guardrail. Otherwise the verdict is INCONCLUSIVE until pending data or pending evaluation metrics are resolved.

formal test & threshold
test:      multi_metric_checklist_canonical_banking_crisis
threshold: MET >= 4 of 5; REFUTE when MET + PENDING_DATA + PENDING_EVAL < 4; refutation guardrail=1

Method

Template
multi_metric_checklist
Clustering
none
Sample
1 countries · 1985 – 2005
Evidence type
canonical_case_multi_metric

Canonical-case checklist evaluator reads canonical_metrics and multi_metric_falsification; no regression model is estimated. Each metric is scored against its pre-registered source, window, and threshold before applying the count rule below.

Data

VariableSourceTransform
real_gdp
outcome
world_bank_wdi:NY.GDP.MKTP.KDtier 2
average_yoy
real_house_price_index
outcome
bis:WS_SPPtier 2
peak_to_trough

● ready  · ● pending  · ● reconstruct-needed

Detailed result card

Result card — banking_crisis_japan_1990_lost_decade

Verdict: supported

Reason: 4 of 5 metrics met threshold (support threshold 4)

Pre-registered rule: SUPPORT if >= 4 of 5 metrics met; REFUTE if <= 1 met (impossible to hit support).

Counts: 4 MET · 0 NOT_MET · 1 PENDING_DATA · 0 PENDING_EVAL

Primary country: JPN

Metric-by-metric

| # | Metric | Status | Observed | Threshold | Notes | |---|---|:---:|---:|---|---| | 1 | equity_index_decline | MET | 72.7 (2003) [peak_to_trough_pct_decline] | >= 60% peak-to-trough decline (Dec-1989 to Apr-2003) | | | 2 | real_house_price_decline | MET | 46 (2005) [peak_to_trough_pct_decline] | >= 40% decline | | | 3 | real_gdp_growth_lost_decade | MET | 0.896 (2002) [average_annual_growth_rate_value] | <= 1% average annual growth | average annual growth 1992-2002 = 0.896; threshold <=1 | | 4 | laeven_valencia_systemic_banking_crisis | MET | 1 (1997) [coded_yes_indicator_max] | coded yes in 1997-2001 (Laeven-Valencia codes Japan 1997-2001) | coded YES evaluated from binary event indicator | | 5 | bank_npl_ratio_peak | PENDING_DATA | | >= 8% of gross loans | No JPN observations in window 1998-2003 |

Claim

Japan's 1990-2003 banking-and-asset-bubble bust — Nikkei index decline of >= 60% peak-to-trough, real residential property price decline of >= 40%, persistent bank-NPL accumulation that required government recapitalisation in 1998 and 2003, and a "lost decade" of GDP growth averaging <= 1% — is a canonical case of a delayed-resolution banking crisis. The hypothesis is that Japan 1990-2003 meets the canonical multi-metric checklist on at least 4 of 5 metrics.

Interpretation

The canonical-case pattern match is satisfied: 4 of 5 pre-registered metrics meet their thresholds, above the support threshold of 4. Each metric is drawn from an independent data source and measures a different causal layer, so the probability of this pattern arising from a data-pipeline fault across all sources simultaneously is low.

Steelman live concerns

See hypotheses/steelman/banking_crisis_japan_1990_lost_decade.md for the strongest opposing arguments. Canonical-case multi-metric evidence is a pattern match, not a causal identification — the result card should be read as 'outcome trajectory matches the predicted pattern to degree X' rather than 'policy P caused the outcome'.

Provenance

Vintages pinned in manifest.yaml. Full per-metric diagnostics in diagnostics.json. Machine-readable results in metric_results.parquet.

Strongest opposing argument

Every hypothesis ships with its charitable opposing argument. The framework earns credibility by handling objections at their strongest, not weakest.

Authored framework. Read the transparency note.