Pre-registration
Monetary base expansion (M2 growth) correlates with asset price inflation in equities and real estate with a lag, measurable via cointegration and lead-lag analysis across major developed economies 2008-2025. The relationship is asymmetric across asset classes and does not require a strict Austrian Business Cycle mechanism to hold.
Falsification criterion — what would disprove this
This hypothesis is considered falsified if:
PRIMARY (dispositive): the hypothesis is SUPPORTED if BOTH asset classes (real equity total return AND real housing total return) show (a) at least 6 of 10 spec countries with a non-negative Pearson correlation between M2-growth(t-1) and real-asset-return(t), AND (b) a panel mean lag-1 correlation > 0. PARTIAL if exactly one asset class clears (consistent with the spec's "asymmetric across asset classes" clause but not the symmetric reading). REFUTED if both asset classes fail both legs. INFORMATIVE: contemporaneous correlations and the sign distribution across countries are reported in diagnostics for transparency. METHOD_VALID: JST Macrohistory must contain money + asset-return series for at least 8 of 10 spec countries with >=5 annual observations each in 2008-2020.
formal test & threshold
test: panel_lag1_correlation_dual_asset_class threshold: PRIMARY: For each asset class, n_countries_with_nonneg_lag1_corr >= 6 AND panel_mean_lag1_corr > 0. SUPPORTED iff both asset classes pass; PARTIAL iff exactly one passes; REFUTED iff neither passes.
Method
- Template
cointegration_vecm- Fixed effects
country- Clustering
country- Sample
- 10 countries · 2008 – 2025
- Evidence type
- associational
Johansen test for cointegration rank; VECM with country fixed effects; lead-lag analysis via cross-correlation functions reported alongside. Robustness: panel-ARDL as secondary specification.
Data
| Variable | Source | Transform |
|---|---|---|
equity_index_real_returns outcome | oecd:KEItier 2 | real_yoy_pct_change_base_2015 |
real_estate_price_index_real outcome | bis:WS_SPPtier 2 | real_yoy_pct_change_base_2015 |
cpi_core outcome | oecd:OECD.SDD.TPStier 2 | yoy_pct_change |
m2_growth_annualised treatment | fred:M2SL (USA)tier 1 ecb:BSI (EU)tier 1 boe:LPMAUYM (GBR)tier 1 boj:MA (JPN)tier 1 rba:D3 (AUS)tier 1 statcan:v41552796 (CAN)tier 1 | yoy_pct_change |
● ready · ● pending · ● reconstruct-needed
Detailed result card
M2 expansion correlates with asset price inflation
Verdict: partial — Housing leg met the >=6/10 + positive-mean thresholds but equities did not. Equities: 5/9 non-negative, mean lag-1 corr = +0.198. Housing: 6/10 non-negative, mean = +0.108. Spec's asymmetry clause is consistent with this outcome but the symmetric-claim formulation is not fully supported.
Summary
- 10-country developed-economy annual panel, JST Macrohistory vintage (effective period 2008-2020; spec asks 2008-2025; tail not yet covered).
- Equity leg: 5 of 9 countries with non-negative M2-growth(t-1) → real-equity-return(t) correlation. Panel mean lag-1 correlation: +0.198.
- Housing leg: 6 of 10 countries with non-negative M2-growth(t-1) → real-housing-return(t) correlation. Panel mean lag-1 correlation: +0.108.
- Falsification threshold: ≥6/10 countries with non-negative lag-1 correlation AND positive panel mean, in at least one asset class.
Method
Annual data from the Jorda-Schularick-Taylor Macrohistory database (jst:money export, which carries the full wide JST_R6 row). For each of the 10 spec countries:
- M2 growth = first-difference of log(money).
- Real equity total return = (eq_tr) / (1 + CPI YoY) − 1, where eq_tr is the JST gross-return index ratio.
- Real housing total return = (housing_tr) / (1 + CPI YoY) − 1; for Canada, where JST has no housing_tr, fallback to real YoY of hpnom.
- Lag-1 correlation = Pearson(M2_growth(t-1), real_return(t)) over the 2008-2020 country sub-sample (≥5 obs required).
Falsification rule (sharpened from spec): the symmetric-claim formulation requires both asset classes to clear; the asymmetric clause ("asymmetric across asset classes") is honoured by reporting a partial verdict when only one class clears.
Data
- jst:money (Jorda-Schularick-Taylor Macrohistory wide panel, containing money / eq_tr / housing_tr / hpnom / cpi columns)
Caveats
- JST annual frequency loses the quarterly lead-lag structure the spec's VECM design contemplated. The cointegration_vecm template is downgraded here to a panel-correlation primary because (a) annual frequency over 13 years yields too few observations for stable Johansen ranks per country, and (b) the spec's threshold is operationalisable as a country-share-of-positive-lag-correlations test without estimating a full VECM. A v2 promotion using quarterly BIS WS_SPP property prices + national M3 series could restore the VECM design.
- Sample period 2008-2020 only; 2021-2025 (post-COVID QE peak and subsequent QT) not covered by the JST vintage on disk. The QE-peak years are exactly the test most favourable to the claim, so the absent tail makes the test slightly conservative.
Strongest opposing argument
Every hypothesis ships with its charitable opposing argument. The framework earns credibility by handling objections at their strongest, not weakest.
Notes
Scope decision per mega-spec D.1.5: this hypothesis does NOT extend to claiming ABCT orthodoxy holds, nor that QE necessarily produces CPI inflation. The empirical claim is scoped to asset-price correlation with M2 growth across major developed economies post-2008.