START WITH THE EVIDENCE

Choose the sample, then inspect every match

The workbench loads source-labelled observations, exposes stale and unavailable states, restores a shareable URL and keeps the matched table and sourced CSV beside the charts. Correlation is descriptive: it does not establish causation or forecast price.

Interactive charts require JavaScript. The interpretation, alignment ledger, reading path and citation checklist below remain available without it; no substitute coefficient is generated.

RangeIndicatorRolling window

GoldLoadingNo valid live or browser-cached series.

DollarloadingChecking Federal Reserve broad dollar index via FRED.Open official source

Real yieldloadingChecking Federal Reserve TIPS series via FRED.Open official source

CPIloadingChecking U.S. Bureau of Labor Statistics via FRED.Open official source

M2loadingChecking Federal Reserve M2 series via FRED.Open official source

Fed assetsloadingChecking Federal Reserve H.4.1 / WALCL via FRED.Open official source

CFTC netloadingChecking U.S. Commodity Futures Trading Commission.Open official source

SilverloadingChecking World Bank Commodity Price Data (The Pink Sheet).Open official source

Loading matched gold and macro observations…

CORRELATION MATRIX

How the available series moved in this selected sample

Gold

Gold1.00

Matrix values use matched observations and descriptive changes. A high or low coefficient is not a causal estimate, a fair value, or a price forecast.

METHOD AND LIMITS

Official observations and GoldObserve calculations stay separate

Official-source layer: the displayed gold, silver, Federal Reserve, BLS/FRED and CFTC observations retain their own observation dates and source labels. World Bank silver is a licensed monthly composite series with historical definition changes, not a continuous modern LBMA daily benchmark.

GoldObserve-derived layer: nearest-date matching, day offsets, rebasing, changes, correlations, rolling windows and scatter positions are calculated by GoldObserve. The matching rule is: nearest calendar-date match within the stated tolerance, without reusing a gold observation

Vintage boundary: this explorer currently uses latest available or latest revised observations (`ex_post_latest`). It is not an ALFRED point-in-time backtest and must not be read as what an investor knew on each historical date. CPI and M2 month labels are observation periods, not release dates; CFTC report dates are position dates, not publication timestamps.

Fallback boundary: when live retrieval fails, only this browser's last successfully validated series is used and marked stale. Missing values are not invented.

Interpretation boundary: correlation changes across samples and regimes. It does not prove causation and this explorer is not a trading signal or price prediction.

Data sources and provider statusReview current coverage, freshness and fallback rules.GoldObserve methodologyReview matching, transformations and unavailable-state rules.

THE SHORT ANSWER

Relationships are useful questions, not automatic explanations

By GoldObserve Research / Published August 4, 2026 / Method review updated August 10, 2026

Gold can respond to opportunity cost, currency translation, purchasing-power concerns, liquidity, stress and positioning at the same time. This tool keeps the comparison descriptive: correlation changes across samples, does not establish causation and is not a price forecast.

Default range5 yearsA selected sample, not all history

Default pairGold / real yieldDaily levels matched by date

Rolling window12 observationsOverlapping descriptive windows

StatisticPearson rAssociation, not causation

These are the initial control settings, not a frozen coefficient. The live result depends on the source versions and matched dates returned for the current view.

Editorial scene of a gold pendulum pulled by two different forces while a researcher moves an observation window along a track, showing that measured correlation depends on the selected sample. — Moving the observation window can change the measured relationship while the underlying system remains exposed to several forces. This is an explanation of sample dependence, not a causal model or price forecast. Original GoldObserve editorial illustration.

STATIC ALIGNMENT LEDGER

Each source keeps its own clock before gold dates are matched

A CPI month, a CFTC Tuesday position and a Federal Reserve Wednesday balance sheet are not same-time observations. GoldObserve keeps both original dates, chooses the nearest unused gold observation inside the published tolerance and reports the day offset. It does not silently forward-fill every missing day.

Static method contract for the seven indicator families in the interactive explorer. It defines transformations and tolerances, not a current coefficient.
Indicator	Source	Source clock	Compared change	Maximum match gap
Broad dollar index	Federal Reserve / FRED	Business-day level	Percent change	10 calendar days
10-year TIPS real yield	Federal Reserve / FRED	Business-day percentage points	Arithmetic change	10 calendar days
U.S. CPI	BLS via FRED	Monthly observation period	Percent change	45 calendar days
U.S. M2 money stock	Federal Reserve / FRED	Monthly level	Percent change	45 calendar days
Federal Reserve total assets	Federal Reserve H.4.1 / FRED	Weekly Wednesday level	Percent change	14 calendar days
CFTC managed-money net	U.S. CFTC	Weekly Tuesday position	Arithmetic change	14 calendar days
Silver monthly price	World Bank Pink Sheet	Monthly average	Percent change	45 calendar days

Monthly series need wider tolerances because their labels identify observation periods, not release timestamps. Wider tolerance does not mean interpolation: unmatched observations are rejected, and the table and CSV preserve both dates for audit.

READ THE EVIDENCE IN ORDER

One coefficient is the start of the investigation

1 · SAMPLEStart with the matrix

Identify which series are actually available. Different cells can use different matched dates and sample sizes, so compare coverage before comparing strength.

2 · STABILITYThen inspect the rolling line

A full-period coefficient can hide sign changes. Look for persistence, breaks and windows dominated by a small number of observations.

3 · SHAPEFinish with the scatter

Check whether the relationship is broad or driven by outliers, clusters or one regime. A straight-line coefficient does not control for other variables.

BEFORE YOU CITE A RESULT

Carry the sample, clocks and revision policy with the number

Which sample?: Record range, selected indicator and rolling window.
Which dates?: Keep the gold date, indicator date and day offset together.
Which versions?: Use the displayed observation time and version hash where available.
Which claim?: Describe association only; do not convert it into causation or a forecast.

Gold Correlation Explorer