Source-labelled market dataSource, observation time and freshness are shown with every quoteData statusAll pages
GoldObserve

MONTHLY LOG RETURNS / ROLLING WINDOWS / NOT A FORECAST

Gold Price Volatility History

Compare rolling 3- to 60-month gold volatility using an empirical volatility cone, reproducible monthly-return formulas and source-labelled local history.

THE SHORT ANSWER

Gold volatility changes through time, and the selected window changes the answer

Rolling volatility measures the dispersion of monthly gold returns inside a moving window. A 12-month window reacts quickly but can be dominated by one shock; 36 and 60 months respond more slowly and are better for regime context. None of them predicts direction. The volatility cone below shows where each latest estimate sits inside its own complete historical distribution.

12-MONTH VOLATILITY+21.0%

Annualized sample standard deviation through 2026-07.

36-MONTH VOLATILITY+14.2%

Annualized sample standard deviation through 2026-07.

60-MONTH VOLATILITY+13.2%

Annualized sample standard deviation through 2026-07.

SAMPLINGMonthly log returns

Not a daily realized-volatility measure.

INTERACTIVE VOLATILITY CONE

Where does today's estimate sit inside each window's own history?

Compare six lookback windows without hiding their historical spread. The shaded bands are empirical percentiles, the line marks each historical median and the points show the latest annualized estimate. Select a window to reveal its exact sample and current rank.

CURRENT POSITION IN HISTORY

Latest 12-month volatility is near the high end of its own history

The latest value is compared only with completed 12-month windows using the same monthly method.

12-month latest21.0%88.8% empirical rank · 2026-07
5th–95th percentile25th–75th percentileHistorical medianLatest value
0%9%18%26%35%3M6M12M24M36M60M
Window12 months
Latest21.0%
Historical middle 50%8.2%14.2%
Empirical rank88.8%

How to read it: wider bands mean that the same lookback window has produced a broader range of historical volatility estimates. The outer band excludes the most extreme 10% of completed windows; those extremes remain in the table and CSV.

Empirical distributions of annualized volatility from overlapping monthly windows. Percentile ranks describe this historical sample, not forecast probability.
WindowSample5th–95th25th–75thMedianLatestLatest rankObserved extremes
3 months7961960-04 onward0.0%27.0%4.6%13.7%8.7%9.8%55.9%0.0%102.8%
6 months7931960-07 onward0.0%27.4%7.0%14.4%10.2%15.3%78.3%0.0%70.2%
12 months7871961-01 onward0.0%28.9%8.2%14.2%10.8%21.0%88.8%0.0%52.5%
24 months7751962-01 onward0.0%27.9%9.2%15.3%11.4%16.0%79.0%0.0%42.1%
36 months7631963-01 onward0.0%27.8%9.8%15.1%11.5%14.2%71.6%0.0%37.0%
60 months7391965-01 onward1.3%29.1%10.2%16.5%12.1%13.2%63.5%0.0%32.3%

ANNUAL VOLATILITY STATE TRANSITIONS

Year-end volatility often persisted, but every starting state also changed

Classify each December's trailing 12-month volatility into lower, middle or upper historical thirds, then follow its state one year forward. The fixed 64-transition denominator and exact years prevent the cell shading from being mistaken for a forecast or a natural risk threshold.

YEAR-END 12M VOLATILITY STATES

Did a low, middle or high state remain in the same historical third one year later?

Outlined: same stateSelected cellDarker: larger row share
ANNUAL TRANSITIONS64

19611962 through 20242025.

SAME HISTORICAL THIRD34

53.1% of descriptive transitions.

LATEST COMPLETE TRANSITIONMiddle thirdMiddle third

20242025: 10.5% to 12.0%.

START ↓ / NEXT →
Lower third
Middle third
Upper third
Lower third
Middle third
Upper third
Period20242025
State transitionMiddle thirdMiddle third
Starting 12M volatility10.5%
Next-year 12M volatility12.0%
Change+1.5 pp

How to read it: each row begins with one historical volatility third; each column shows the state at the following December. Counts and percentages use the row's starting states as the denominator. The thresholds are lower ≤ 9.3%, middle ≤ 12.5% and upper above 12.5% in this retained sample.

Year-end 12-month volatility state transitions. Percentages use the starting-state row as denominator and are historical descriptions, not transition probabilities.
Starting stateNext-year stateTransitionsStarting-state totalRow shareExact starting years
Lower thirdLower third142263.6%1961, 1962, 1963, 1964, 1965, 1966, 1970, 1991, 1994, 1995, 1996, 1997, 2001, 2017
Lower thirdMiddle third42218.2%1967, 2004, 2018, 2023
Lower thirdUpper third42218.2%1971, 1992, 1998, 2002
Middle thirdLower third22010.0%1969, 2000
Middle thirdMiddle third102050.0%1968, 1984, 1987, 1988, 2012, 2013, 2014, 2019, 2020, 2024
Middle thirdUpper third82040.0%1977, 1985, 1989, 2005, 2007, 2010, 2015, 2021
Upper thirdLower third52222.7%1990, 1993, 2003, 2016, 2022
Upper thirdMiddle third72231.8%1976, 1983, 1986, 1999, 2006, 2009, 2011
Upper thirdUpper third102245.5%1972, 1973, 1974, 1975, 1978, 1979, 1980, 1981, 1982, 2008

Source: World Bank monthly-average nominal USD gold. Volatility states and transitions are GoldObserve calculations from the unchanged locally versioned series.

INTERACTIVE ROLLING WINDOW

Compare fast and slow measures without treating either as a forecast

Choose 3, 6, 12, 24, 36 or 60 months, then inspect the exact source-derived value with pointer, touch or keyboard controls. The chart annualizes monthly dispersion with square-root-of-time scaling.

LOCAL LICENSED HISTORY

36-month annualized volatility

FIRST OBSERVATION1963-01+0.00%
LATEST OBSERVATION2026-07+14.25%
LOWEST+0.00%Inside the selected sample
HIGHEST+37.00%Inside the selected sample
1963-011994-102026-07
Observation2026-07
36-month annualized volatility+14.25%
Series basisMonthly averages

Crosshairs snap to a calculated observation. Monthly averages smooth intramonth highs, lows and drawdowns; this chart is not a daily close series. Historical windows overlap and are not independent forecasts.

This export contains GoldObserve-derived or source-cleared observations with citation metadata.

Source: World Bank Commodity Price Data (The Pink Sheet), monthly nominal USD per troy ounce, CC BY 4.0. GoldObserve serves the bundled local series; page loads do not download the upstream workbook.

WINDOW COMPARISON

Latest, minimum and maximum annualized monthly volatility

Statistics use rolling monthly log returns and sample standard deviation. Maxima and minima are historical sample results, not risk limits.
WindowFirst availableLatestLatest volatilityHistorical minimumHistorical maximumObservations
3 months1960-042026-07+9.8%+0.0%+102.8%796
6 months1960-072026-07+15.3%+0.0%+70.2%793
12 months1961-012026-07+21.0%+0.0%+52.5%787
24 months1962-012026-07+16.0%+0.0%+42.1%775
36 months1963-012026-07+14.2%+0.0%+37.0%763
60 months1965-012026-07+13.2%+0.0%+32.3%739

REPRODUCIBLE FORMULA

Monthly log returns, sample dispersion and square-root-of-time scaling

01Calculate returns

r(t) = ln[monthly average(t) / monthly average(t-1)].

02Select a window

Keep the latest 3, 6, 12, 24, 36 or 60 monthly log returns.

03Measure dispersion

Calculate sample standard deviation using n - 1 in the denominator.

04Annualize

Multiply monthly standard deviation by sqrt(12), then express it as a percentage.

Square-root-of-time annualization is a convention, not proof that returns are independent or normally distributed. Serial correlation, volatility clustering and jumps can make realized outcomes differ from a simple scaled estimate.

HOW TO USE THE NUMBER

Volatility is a risk description, not a buy or sell signal

PHYSICAL BUYERShort holding periods combine market dispersion with a large spread.

A monthly volatility estimate does not include premiums, dealer bids or time needed to liquidate.

PORTFOLIO ALLOCATORUse matched return frequencies across assets.

Comparing monthly gold volatility with daily equity volatility without reconciliation is misleading.

TRADERA backward-looking window can change after the move.

High realized volatility may persist or collapse; it does not reveal direction.

RESEARCHERRecord currency, sampling and annualization.

A USD monthly-average result is not interchangeable with local-currency or daily-close volatility.

Source and statistical limitsWorld Bank monthly history · CC BY 4.0

SOURCE, LICENSE & REPRODUCIBILITY

The statistics use one consistent monthly series

The source is the World Bank Commodity Price Data (The Pink Sheet), licensed CC BY 4.0. The workbook was published 2026-08-04, retrieved 2026-08-07, and recorded with SHA-256 7902a77505ebdc5d202ce65f666c2ee1b04b626f042d7738ed3e6f7d112c8433.

The values are monthly averages in nominal US dollars per troy ounce. They are not London auction prices, daily closes, intraday highs or executable dealer quotes.

CONTINUE THE RESEARCH

Keep monthly long-run statistics separate from recent daily data

FREQUENTLY ASKED QUESTIONS

Questions about this historical statistic

How is historical gold volatility calculated here?

GoldObserve calculates monthly logarithmic returns, takes the sample standard deviation inside a rolling window and annualizes it by multiplying by the square root of 12.

Why are 12-, 36- and 60-month volatility different?

Short windows react faster to recent changes. Longer windows are smoother and retain older observations, so they answer different questions.

Does volatility show whether gold went up or down?

No. Volatility measures dispersion, not direction. A rising and a falling period can have the same volatility.

Is monthly volatility the same as daily volatility?

No. Sampling frequency changes the return series and can hide intramonth movement. Do not compare the figures without naming the frequency and method.

Can high volatility forecast a gold-price reversal?

No. It describes how variable recent monthly returns were. It does not identify direction, fair value or the timing of a reversal.