MEASURED · 5 OCTOBER 2026

Is silver a good investment? What buying silver in every month since 1971 returned

Arguments about silver are easy to find. This page counts instead: it treats each of the 661 months from August 1971 to August 2026 as a day someone bought silver, and records what happened to them after 1, 5, 10 and 20 years, next to a gold buyer in the same month. Then it checks whether conditions like today’s told buyers anything.

The short answer

Silver has made money over long holds more often than not, but less often than gold, and with much deeper holes on the way. After inflation, the median 10-year silver buyer since 1971 made −1.31% a year, and since 1968 silver has fallen by half or more from a peak 8 times.

How often silver buyers were up, next to gold

Every start month with a full holding period, monthly average prices; “typical year” is the median yearly return, “real” is after US consumer-price inflation. Gold columns in amber. Before costs and storage.

The gaps are widest where most people actually hold. Over a single year, silver lost 20% or more in 12.2% of start months, gold in 5.5%. Over five years, 24.3% against 15.6%.

Silver against gold, head to head

Silver’s wins come in bursts: its three biggest 12-month leads over gold were for buyers in late 1978 to early 1979, late 2009 to 2010, and December 2024 to 2025. Held 10 years or more, it came out ahead in only a quarter to a third of start months. Whether the gold-silver ratio tells you when a burst is coming is tested in the gold-silver ratio, tested.

How long buyers waited to get their money back

From each start month: the wait until the monthly average first got back to the price paid. 84.4% of buyers were back within a year and 94.7% within ten. The rest waited a long time. Buyers in January 1980 waited 375 months, until April 2011; 31 start months (4.7%) waited more than ten years. After inflation it is worse: silver’s January 1980 average is worth $167.49 in August 2026 dollars, and in August 2026 silver was 61% below it. Gold, by contrast, was at a new inflation-adjusted record in February 2026.

Is now a good time to buy silver?

We marked six kinds of month people ask about and compared what followed with all months. Each result was checked against 2,000 random histories: silver’s real price rebuilt from its own monthly moves in shuffled 12-month blocks, so they look like silver but cannot know the future. In August 2026, 3 of the six were true: up 30% or more over 12 months, 20% or more below the record and real price in its top fifth.

“Chance matches it” is for the next-year result: how often random histories produced a gap from the baseline at least as large in the same direction. Stretches are runs of qualifying months merged when under a year apart. A result can be rare by chance and still rest on very few stretches.

After a 30% year, nothing special. This is true now. Following a 12-month rise of 30% or more, silver was up a year later 52.5% of the time, the same as any month (52.5%), across 13 separate stretches. For gold, a big year was followed by a better-than-chance next year; for silver it was not.

New records were followed by strong years, but rarely. Months at a new monthly record were up a year later 88.4% of the time, with a typical gain of 62%, and random histories matched that only 0.2% of the time. But those months fall into just 3 stretches with a year of results, 1971, 1978, 2011, and the single record month of April 2011 was followed by a −26% year. The record fits momentum; it does not prove it.

An expensive real price was followed by losses, but chance does the same. Also true now. After months with silver’s inflation-adjusted price in its top fifth, the next five years beat inflation only 17.1% of the time. Random histories produced a gap that large 25.2% of the time, because in any record the highest stretch tends to be followed by lower ones. That is not evidence that a high price predicts a fall. Being 29% below the record, also true now, told buyers nothing either (8.5% by chance, 3 stretches).

All at once or month by month?

Spreading a purchase over a year did not raise the typical result; over one year the lump sum won 52.5% of the time. What it did was soften the worst case, from −63% to −40% over a year, because no single month’s price decided everything.

What buying and selling costs do

The tables above ignore costs. Coins and bars are bought above the silver price and sold below it. With a round-trip cost of 2% of the money, a 1-year hold was up 50.4% of the time instead of 52.5%; at 6%, 45.1%. Over 10 years, 62.7% with a 6% cost. The IRS also taxes a gain on silver held over a year as a collectible, at up to 28%. What a bar or coin costs over the metal is in how much is a silver bar worth?

What the record says, in practice

• Deep falls are normal for silver. Its worst 5-year result was −84%, and a 1-year hold lost 20% or more in 12.2% of start months.
• Gold did the same job more reliably. Over every holding period, gold was up and beat inflation more often.
• No timing signal survived the check. Records, big years, falls and “cheap” or “expensive” real prices were either matched by chance or rest on a handful of stretches.
• Buying monthly does not raise the return. It limits how much one bad month can hurt.

The year-by-year record is in silver price history, the same study for gold in is gold a good investment?, and what has moved silver in why is silver going up? This page is a measurement of the past, not a forecast or personal advice.

How we measured it

Prices: World Bank Commodity Price Data (the Pink Sheet), monthly average silver and gold prices in US dollars per troy ounce, release of 2 October 2026. Inflation: BLS CPI-U, not seasonally adjusted; October 2025, never published, is filled from its neighbours. Window: August 1971, when the gold price floated, to August 2026, the last month with both a price and a CPI. Holds: every start month with a full period, monthly averages, no costs unless stated. Chance check: 2,000 histories of silver’s real price rebuilt from its own monthly log changes in 12-month blocks; the nominal path is that real path times the actual CPI.

Common questions