Is market conductance a forecastable state variable?
A falsification study of the price impact operator in gold and nine other markets. The foundational claim survives everywhere. The mechanism proposed to explain it does not.
Abstract
Most market forecasting targets the price. This study targets the operator that
converts news into price movement — written G in
ΔP = G · ε, where ε is the shock and
G is the market's conductance. Two questions were pre-registered: is
G a genuine state variable rather than a constant with noise, and is it explained by
the discretion distribution of the holder base — who is forced to act, when,
and who is left to absorb them?
The first is true and universal. The second is not supported. Conductance is strongly persistent and forecastable in every liquid market tested, at p < 0.0001. But a signed conductance measure built from holder-base data fails three pre-registered kill conditions, is essentially uncorrelated with what it was built to forecast (corr = 0.014), and adds nothing on scheduled policy events where its predicted effect would be largest.
Research engineering and infrastructure by ITMaster and Code Nest.
Findings at a glance
| Test | Result | Verdict |
|---|---|---|
Is G a state variable, or a constant with noise? | Ljung-Box(22) = 1021.5 vs null p99 = 40.0; out-of-sample R² +7.3% | STATE VARIABLE |
| Does holder-base Λ beat realised + implied volatility? | Out-of-sample R² falls in 5 of 6 specifications; corr(log Λ, log G) = 0.014 | FAILED |
| Does signed asymmetry produce signed drift? | Correct sign at every horizon, |t| ≤ 1.09 | FAILED |
| Do large moves cluster in high-Λ states? | Non-monotone; high/low quintile ratio 0.68 | FAILED |
On scheduled events, does pre-event G predict the response? | Coefficient −0.123; predicted +1.0 excluded at 5.4 standard errors | FAILED |
| Does the conductance result generalise beyond gold? | Identical signature in all 10 assets, p < 0.0001 | UNIVERSAL |
Read the full results, with every coefficient and confidence interval →
Conductance is real — and it is everywhere
The same measurement, run unchanged across ten liquid instruments. Every one rejects the constant-impact null decisively, and gold sits mid-pack rather than standing out.
| Ticker | Asset class | α̂ | Ljung-Box(22) | VR(66) | OOS R² |
|---|---|---|---|---|---|
| HYG | High-yield credit | 0.994 | 1632.6 | 7.21 | +10.1% |
| SLV | Silver | 0.797 | 2459.0 | 11.02 | +9.4% |
| SPY | US large-cap equity | 1.357 | 1208.2 | 8.01 | +8.7% |
| QQQ | US tech equity | 1.022 | 1134.6 | 7.68 | +8.1% |
| USO | Crude oil | 0.734 | 964.4 | 6.30 | +7.3% |
| GLD | Gold | 1.032 | 1021.5 | 6.82 | +7.3% |
| EEM | EM equity | 1.257 | 1103.5 | 7.45 | +7.3% |
| GDX | Gold miners | 1.240 | 1173.6 | 7.45 | +7.2% |
| TLT | Long Treasuries | 1.027 | 729.5 | 6.46 | +5.6% |
| FXE | Euro | 0.412 | 777.0 | 6.51 | +5.2% |
A mechanism built on gold's uniquely documented holder base cannot explain a signature that appears just as strongly in high-yield credit and the euro.
Why the holder-base explanation fails
Two independent tests point the same way. First, every Commitments of Traders variable tested individually: Managed Money positioning — the theory's proxy for margin-sensitive, least-discretionary capital — does not forecast conductance at any horizon, with every t-statistic below 1.2 in absolute value. What does forecast it is generic market depth and breadth, with the negative sign absorption implies. That is textbook microstructure, not a new mechanism.
Second, the scheduled-event test. FOMC decision dates are fixed years ahead, so news arrival carries no information and the size of each policy surprise is close to unpredictable. On 137 decisions, pre-event conductance carried a coefficient of −0.123 (t = −0.59) while option-implied volatility priced the same risk at 1.005 (t = +2.75). The theory's predicted +1.0 sits 5.4 standard errors outside the confidence interval.