What the winner over your window means
The backtest runs gold against the S&P 500 with dividends reinvested and reports which ended ahead. Both answers are accurate for the window they cover, and neither is accurate for a window a few years either side of it.
- Gold ahead The metal won this window
- Windows opening near a market top or a gold trough produce this, and March 2000 is the clearest case, since it starts at the peak of the dot-com bubble with gold near a twenty-year low. The finding is real, and it says as much about where the window opened as about either asset.
- The S&P 500 ahead Equities won this window
- Long windows that include the 1980s and 1990s generally land here, because gold lost money outright over those two decades while equities compounded with dividends on top. That is the other half of the record and it is the half a chart selected to make a point tends to leave out.
Why the start date decides this argument
Almost every claim about gold beating the stock market is accurate for the window it names and stops being accurate a few years either side of it. That is not an accusation of dishonesty. It is what happens when two assets have both had decade-long stretches of leading and lagging, and somebody has to pick a start date to make a comparison at all.
March 2000 is the window you will usually be shown. It begins at the top of the dot-com bubble and at a gold price near a twenty-year low, so it stacks both legs in the same direction before the comparison starts. January 1980 to December 1999 is the window nobody shows you, and over those twenty years gold lost money outright while the S&P 500 had one of its strongest runs on record.
One detail decides the long-run answer, and most published versions of this comparison get it wrong. Dividends are worth roughly three percentage points a year on the S&P 500, which compounds to more than the entire gap between the two assets over 58 years. Gold produces no income at all, so its price change is its whole return, and setting it against an equity price index rather than a total-return one is not a small omission, which is why this backtest reinvests them.
What gold and stocks each do in a portfolio
Stocks are ownership of businesses that earn, reinvest and pay dividends, and over the 58 years in our series the S&P 500 with dividends reinvested compounded at about 10.8% a year against gold's 8.5%. Equities are the growth engine of a retirement portfolio and nothing here suggests otherwise, since roughly three percentage points a year of that equity return is dividends that gold has no equivalent of.
Gold is not competing for that job. It produces no income at all, which is the honest cost of holding it, and in exchange it is priced by forces that have little to do with corporate earnings. That is why the last twenty-five years run the other way, with gold compounding at about 11.6% a year from the end of 2000 against 9.0% for the S&P 500 on the same total-return basis, across a window containing two equity bear markets.
The reason to run this comparison in both directions is that the start date decides the answer more than either asset does. March 2000 opens at the top of the dot-com bubble with gold near a twenty-year low, and January 1980 opens at a gold spike with two decades of equity expansion ahead. A claim that survives a five-year shift in the start date is worth more than one that does not, and this tool exists so you can apply that test yourself.
How to run the backtest
Set the window you were quoted, then move the start date five years and watch what happens.
-
Set the window you were quoted
If you have seen a period quoted where gold beat stocks, put that window in first. Most of those figures are accurate for the window they name, so the useful move is to widen it by a few years and watch what happens to the answer.
-
Then move the start date by five years
This is the whole point of letting you pick. A window starting in 2000 and a window starting in 1995 produce opposite answers on the same two assets, and the difference is entirely the start date.
-
Read the annualised figure, not the total
A total return across twenty years is not comparable to one across seven. The annualised number puts two windows of different lengths on the same footing, which is the only way to tell whether a period was genuinely better or simply longer.
-
Check the worst fall alongside the ending value
Both assets have had falls deep enough to end a holding period early. A window where gold finished ahead having fallen 60% on the way is not the same product as one where it finished ahead having fallen 20%, and the table under the chart carries both numbers.
What each input means
The two dates are the whole experiment, and the starting amount only scales the output.
- Start year
-
This is the month the comparison begins. The gold series starts in April 1968 and the equity series runs back much further, so the window is bounded by gold rather than by stocks.
Where to find it The year in whatever claim you are testing. If none was given, try 1968, 1980 and 2000 and see how far the answers move.
- End year
-
Where the window closes, defaulting to the most recent month-end. Ending a window on a market low rather than at the present is the other half of how a backtest gets steered.
Where to find it Today, unless you are checking a specific historical claim.
- Starting amount
-
The sum to run through both series. It scales the dollar outputs and leaves every percentage untouched.
Where to find it The amount you are weighing up, or leave the default.
How the comparison is calculated
Both legs run on month-end closes, and the limitation on the equity side is stated rather than patched over.
Both legs start with the same amount on the same date and run on annual closes. The gold series is our own record from 1968. The equity leg is an S&P 500 total-return series with dividends reinvested, which is the honest basis for a comparison against an asset that pays no income at all. Annualised change is the compound rate between the two endpoints, and the drawdown is the deepest peak-to-trough fall inside the window.
annualised = (ending value ÷ starting amount) ^ (1 ÷ years) − 1
- ending value
- What the starting amount grew to by the end of the window
- years
- Length of the window in years
- drawdown
- Largest fall from a running peak to a subsequent trough inside the window
What this comparison leaves out
- Every cost of owning bullion in a gold IRA, from the setup fee and the custodian's annual charge to the storage fee and the 5% to 8% markup over spot paid on both sides of the trade.
- Fund fees and taxes on the equity side, which work in the other direction.
- Anything that happened inside a year, because this runs on annual closes and cannot see between two of them.
- Any forecast, because four price regimes in 58 years is a small sample and this describes them rather than extending them.
Three windows, three different answers
The window a dealer will quote, the window nobody quotes, and the longest one the data supports.
The window a dealer will quote
March 2000 to the present, starting at the dot-com peak.
- Start
- 2000
- End
- 2026
- Amount
- $100,000
Gold ahead, even with dividends counted on the equity side.
This window is accurate and it is chosen. It begins at the highest equity valuation in a generation and at a gold price near a twenty-year low, so it stacks both legs in the same direction before the comparison starts.
The window nobody quotes
January 1980 to December 1999, starting at the gold peak.
- Start
- 1980
- End
- 1999
- Amount
- $100,000
Gold down heavily in nominal terms while the S&P 500 compounded.
Twenty years in which gold lost money outright and equities had one of their strongest runs on record. It is the same asset and the same country, and only the start date changed.
The full record
April 1968 to August 2026, the longest window the data supports.
- Start
- 1968
- End
- 2026
- Amount
- $100,000
The S&P 500 ahead, once dividends are counted.
The honest version of the long-run comparison, and the one that rarely appears in gold marketing. Against a price index gold wins this window. Against the total-return index it does not, and the difference between those two answers is roughly three percentage points a year compounded over 58 years.
The record both sides are drawn from
Four figures explain why the same two assets produce opposite answers depending on where you start.
- 697 months The comparable record
April 1968 to August 2026, bounded by when the London gold price began to float rather than by the equity data, which runs to 1871.
- About 3 points What dividends add, annually
The difference between the S&P 500 price index and its total-return version across the long run, and the reason this backtest uses the total-return series. Over 58 years compounded, that gap is larger than the entire gap between gold and equities.
- Zero Income gold produces
No dividend, no coupon, no rent. Its price change is its whole return, which is why the gold leg needs no adjustment and the equity leg does.
- 1980 to 1999 The twenty years gold lost money
Nominal decline across two decades while inflation ran positive throughout. Any backtest that starts after 1999 excludes this period entirely.
What people get wrong about gold versus stocks
The first is about window selection and the rest are about what a backtest can carry.
Has gold beaten the stock market since 2000?
Yes, and the window is doing most of the work. March 2000 was the top of the dot-com bubble and gold was near a twenty-year low, so the comparison starts with one asset at its most expensive and the other at its cheapest. Move the start to 1995 and the answer flips.
Is a price index close enough for this comparison?
No, it is short by about three percentage points a year, which compounds to more than the entire difference between the two assets over the long run. Most published gold-versus-stocks charts use one anyway, because it flatters the gold side, so it is worth checking which index any comparison you are shown is actually built on.
Does the backtest show what a gold IRA would have returned?
No, it shows what the metal did. A real account pays a setup fee, annual administration, storage and a 5% to 8% markup on the way in, then a dealer's bid discount on the way out. That round trip has to be earned back before any of these figures start applying to you.
Is sixty years of data enough to settle the argument?
No, sixty years is about one gold cycle and a bit. The metal had a fifteen-year rise, a twenty-year fall, a twelve-year rise and another decade of chop, so the record contains about four regimes and a backtest is a sample of them rather than a law.
Gold vs stocks questions
Has gold outperformed the S&P 500?
It depends entirely on the window, and that is not a dodge. Since March 2000 gold is well ahead of the S&P 500 price index. From January 1980 to December 1999 it lost money outright while equities compounded. Over the full record since 1968 gold beats the price index and loses to the total-return index once dividends are counted.
Does this include dividends?
Yes. The equity leg is a total-return series with dividends reinvested, which matters more than people expect: about three percentage points a year, compounding to more than the entire gap between the two assets over the full record. Gold produces no income, so setting it against an equity price index would be comparing one asset's whole return against another's partial one.
What is the longest gold has underperformed?
The twenty years from the January 1980 peak to the August 1999 low, during which gold fell in nominal terms and lost roughly 80% of its purchasing power while the S&P 500 had one of its strongest stretches on record. That period is longer than many people's remaining working life at the point they open a gold IRA.
Should I pick gold or stocks for my retirement account?
Neither, and no backtest can answer it for you. What it does is let you check the specific claim you were shown by moving the start date, and a claim that survives a five-year shift in either direction is worth more than one that does not. Our allocation calculator handles the sizing question instead.
Is a 60-year backtest long enough?
No, not for anything you want to call a law. It contains about four distinct regimes for gold, which is a small sample. The record is long enough to show that both assets have had decade-long stretches of leading and lagging, and short enough that fitting a trend to it would be reading more than the data carries.
Where does the gold data come from?
Our own monthly close series, taken at the spot price and recomputed from the publisher's own file, then republished as a CSV you can download and check against these results. It runs from April 1968, which is where the London price begins to float after the collapse of the gold pool, and it is the same series behind every chart on the site.