Every time gold runs, the same question comes back: should you buy the metal itself, or the companies that dig it out of the ground? Gold vs gold miners is an old argument. The case for miners sounds good. When the gold price rises, a miner’s profit rises much faster, and miners pay dividends while a gold bar pays nothing. The case against them is just as simple. A miner is still a company, with rising costs, debt, bad acquisitions, and now and then a bankruptcy.
I wrote this to settle it with real prices instead of opinions, for a regular investor and not a trader. The plan is one most of us can actually follow: put $1,000 in every quarter (January, April, July, October), never sell, reinvest any dividends. I started in April 1985, as far back as I have clean prices for the big US-listed miners, and ran the same plan in gold, in four big miners, in one or five miners picked at random, and in the miner ETFs GDX, GDXJ, and RING. Those funds hold only mining shares, no bullion. For each one I show what the money grew to, the return per year, and the worst fall you would have had to sit through.
The answer turned out to depend on when you stop counting. Valued at the end of 2024, gold beat the miners in every version of the test. Valued in October 2026, most miner versions beat gold. The only thing that changed in between was a 22-month miner rally. So the short version is this: miners are not a better gold. They are gold with leverage, and the result swings with the month you look at the account.
Gold, 41 years
$1.24M
$167,000 paid in, 8.0% a year
Four big miners
$1.51M
8.7% a year, dividends reinvested
Same plans at Dec 2024
$603k
miners, against $791k in gold
Worst drop
-76%
miner basket. Gold: -47%
Why miners should beat gold, and why they usually do not
The textbook case for miners is operating leverage. A mine with an all-in cost of $1,400 an ounce earns $600 at a $2,000 gold price. Gold goes to $2,400 and the margin goes to $1,000. A 20% move in the metal becomes a 67% move in profit. Add dividends on top, and the stock should run ahead of the bullion it sells.
The leverage works the other way too. Costs rise with energy, wages, and royalties. Mines run out. Managers overpay for the next deposit right when gold is expensive, which is exactly when the shareholders can least afford a write-down. Barrick, Kinross, and plenty of others booked billions in impairments after the 2011 peak. A bar of gold has none of those problems. It also pays nothing.
So the question is not which one is “better”. It is whether the leverage and the dividends pay for the extra damage over the years you actually hold. That is a measurable question, so I measured it.
How I ran the gold vs gold miners test
The plan
$1,000 at every quarter start (January, April, July, October). No selling, no rebalancing. Dividends are reinvested (Yahoo and Tiingo adjusted prices). The account is valued at the last full month, October 2026. I also show the money-weighted return (IRR), because with quarterly buying that is the honest “percent a year” number. CAGR and maximum drawdown are for the instrument itself, as if you had bought once at the start.
Prices on the same footing
The long gold series is the World Bank Pink Sheet monthly average, which goes back much further than any miner. Mixing a monthly average for gold with end-of-month closes for stocks would be unfair either way, so every miner and ETF here is the monthly average of its daily total-return prices too. That is close to what someone buying “sometime during the month” pays. As a check, GLD (the gold ETF) tracks this series within its 0.40% annual fee over 22 years.
What I counted as a gold miner
Gold-first producers listed in the US in dollars, ADRs included. No royalty or streaming companies, no silver-first miners, no explorers with nothing coming out of the ground. The random pool has 28 names, including ones that are gone: Goldcorp, Yamana, Newcrest, Kirkland Lake, Tahoe, Klondex, Richmont, AuRico, New Gold, Golden Star, Pretium, Primero, Timmins Gold, Banro, which went bankrupt and fell more than 99%, and Allied Nevada after its bankruptcy filing. If a stock you hold stops trading, because of a bankruptcy or a takeover, its last price is the exit and the money moves into another random miner that is still trading. ETFs are only the ones that hold miner shares, not bullion: GDX, GDXJ, and RING.
41 years: gold vs four big miners
In 1985 four gold-first miners already had US listings with clean daily data: Newmont, Barrick, Agnico Eagle, and Gold Fields. Here is what $1,000 a quarter did in each, from April 1985 to October 2026, $167,000 in total.
| Quarterly plan | End value | Return on money in | Buy-once CAGR | Worst drop | Volatility |
|---|---|---|---|---|---|
| Gold | $1,235,122 | 8.0% | 6.3% | -47.3% | 12.5% |
| Newmont | $872,798 | 6.8% | 7.1% | -75.0% | 30.5% |
| Barrick | $1,075,142 | 7.5% | 11.9% | -86.9% | 31.2% |
| Gold Fields | $1,335,188 | 8.3% | 5.1% | -86.4% | 38.6% |
| Agnico Eagle | $2,776,451 | 10.8% | 7.6% | -86.4% | 36.1% |
| Four big miners, split evenly | $1,514,895 | 8.7% | 9.4% | -76.2% | 29.6% |
| Same basket, price only | $1,039,831 | n/a | n/a | n/a | n/a |
Two of the four lost to gold. Newmont, the name most people think of first, finished $360,000 behind. Agnico Eagle is the hero of the table, and it is the only one that won by a lot. The basket won by about $280,000, and dividends did all of that work. Without them it ends at $1.04 million, behind gold. The dividends added 46% to the miner basket. Gold gives you nothing to reinvest, so the miners had to win this one on income, not on price.
The “buy-once CAGR” column looks odd next to the plan returns, and that is worth a sentence. Barrick compounded at 11.9% from 1985 because it grew from a small company into one of the biggest producers in the world by the mid-1990s. A quarterly buyer only had a few early contributions in that run. Most of their money went in later, during a long slide. The timing of the money matters more than the CAGR of the stock.
Also note what this table is not. These four are the big miners that survived to 2026. In 1985 a buyer could just as well have picked Homestake, Placer Dome, Echo Bay, Battle Mountain, or Pegasus Gold, which went bankrupt in 1998. I could not find daily prices for those in any free source, so they cannot go into a random draw. What I can do is check the four against the Philadelphia Gold/Silver index (XAU), which held those companies at the time.
The check is not kind. Price only, $1,000 a quarter from April 1985 to December 1999, $59,000 paid in: the XAU plan was worth $41,900, gold $46,300, and the four survivors $89,100. The sector as a whole lost money for 15 years, and the four names that made it look twice as good. Carried on to October 2026, the XAU plan ends at $592,000, against $1.04 million for the four on price alone and $1.24 million in gold. The index pays no dividends in this data and also holds some silver miners, so it understates the sector a bit. Not by enough to close that gap.
This chart is the shape of the whole argument. The miners ran far ahead of gold in the late 1980s and 1990s, mostly on Barrick and Agnico. Then gold went up more than six times from 2001 to 2011, and the miners went up far less. By the end of 2015 the basket’s 30-year lead over gold was almost gone. The Philadelphia Gold/Silver sector index, price only, went from $1.33 for every $1 in gold in 1996 to 16 cents in 2015. That is a sector losing 88% against the metal it digs up, before dividends.
The last 22 months decided gold vs gold miners
Gold averaged $2,648 in December 2024 and $4,105 in October 2026, up 55%. GDX went up 2.4 times over the same months. That is the leverage case working exactly as advertised. But it changes the result of decades-long plans in less than two years.
Valued at the end of 2024, a gold buyer since July 2006 had $155,800. The four big miners had $127,900. GDX had $96,600, a return of 2.8% a year on 18 years of contributions. The median random miner had $107,400. Gold beat every miner plan. Twenty-two months later the order is flipped, and the four big miners finish at $344,900 against $250,300 in gold.
It goes the same way for the longer windows. From 1985, gold led the basket $791,000 to $603,000 at December 2024. From 2000, $357,500 to $239,200. A random single miner from 2000 beat gold in 18% of my simulated plans, at that date. If you wrote this article two years ago, the verdict was “buy gold”. I do not think the right conclusion is that 2026 fixed the miners. The right conclusion is that a miner plan is a much bigger bet on how the cycle ends.
GDX, GDXJ, RING: the miner ETFs against gold
The ETFs answer the “I do not want to pick a stock” version of the question. They only hold miner shares, so their result is a clean read on the sector.
| Quarterly plan, to Oct 2026 | Paid in | Gold plan | ETF plan | ETF buy-once CAGR | ETF worst drop |
|---|---|---|---|---|---|
| GDX, from Jul 2006 | $82,000 | $250,273 | $242,572 | 4.9% | -77.4% |
| GDXJ (juniors), from Jan 2010 | $68,000 | $176,333 | $187,171 | 2.3% | -87.1% |
| RING, from Apr 2012 | $59,000 | $149,243 | $209,327 | 5.3% | -75.0% |
| GLD (gold ETF), from Jul 2006 | $82,000 | $250,273 | $238,282 | 9.2% | -40.5% |
After 20 years and a miner boom, GDX still finished slightly behind gold, $242,600 to $250,300. Its buy-once return was 4.9% a year against 9.7% for the metal, with a drop of 77% along the way. GDXJ, the junior miners, made 2.3% a year buy-once and fell 87% at the worst point. RING started in 2012, near the top of the last gold cycle. That is the only reason it shows such a large win: its quarterly buyers loaded up during the cheap years, 2013 to 2019.
GLD is in the table for the fee. It ended $12,000 behind the World Bank gold series, which is about what 0.40% a year costs over two decades. If you buy gold through an ETF, that is the price of not storing a bar.
One random miner, or five
Picking Agnico Eagle in 1985 would have been great. Nobody knew that in 1985. So I let a random number generator pick: one miner from the names trading at the start, $1,000 into it every quarter. If that company goes bankrupt or gets bought, you take whatever its last price was, and the money plus all later buys go into a new random miner that is still trading. 4,000 runs per window.
The draws start in 2000. Before that the only miners I have prices for are the ones that survived, so a “random” pick in 1985 would just choose among the winners. In January 2000 the pool has 11 names, and it grows to 28 over time, losers included.
| Start, valued Oct 2026 | Gold plan | 1 random miner: 10th pct / median / 90th pct | Beat gold | 5 random miners, median | Beat gold |
|---|---|---|---|---|---|
| Jan 2000 ($108k in) | $562,997 | $347k / $570k / $980k | 52% | $619k | 67% |
| Jul 2006 ($82k in) | $250,273 | $189k / $335k / $598k | 70% | $343k | 93% |
| Jan 2010 ($68k in) | $176,333 | $155k / $253k / $468k | 76% | $280k | 99% |
From 2000, one random miner was a coin flip against gold, with a spread from $347,000 to $980,000 on the same $108,000. With only 11 names at the start, those percentiles land on specific companies, so read them as rough. Five random names lifted the odds to two in three. The later starts look much better because they bought heavily in 2013 to 2019, when miners were cheap against gold, and then rode 2025 and 2026. The typical random single-miner account still fell 60% to 77% from its peak at some point, depending on the start. A run that drew Banro kept buying it all the way down and moved to a new name only when the shares stopped trading in 2018.
Every 10-year window: who won more often
The results so far all end in October 2026, and we just saw how much one end date can move them. Most people do not invest for exactly 41 years either. So I asked a more practical question: if you had run this plan for ten years, starting at any point in the past, how often would the miners have beaten gold?
Here is how it works. Take one start date, say October 2005. Put $1,000 a quarter into gold for ten years, $40,000 in total, and see what it is worth at the end. Do the same with the four big miners. In that case gold ended at $45,500 and the miners at $17,000, less than the money paid in. The miners made 63% less than gold. Then move the start forward one quarter and repeat. From April 1985 to October 2016 that gives 127 separate ten-year plans.
The chart falls into four phases. Plans started from 1985 to 1987 did very well in miners: the 1985 start ended with almost three times as much as gold. Plans started from 1988 to 1998 mostly won, but by modest margins, between 10% less and 48% more than gold. Every plan started from 1999 to 2009 lost to gold, and those started from 2002 to 2006 made 43% to 62% less, because they bought through the long stretch when gold rose and miners did not. Plans started from 2010 won again, because they bought cheap miners in 2013 to 2019 and the later ones finished in the 2025 to 2026 rally.
| Ten-year quarterly plan in | Plans tested | Ended with more than gold | Ended with less than gold |
|---|---|---|---|
| Four big miners, from 1985 | 127 | 77 (61%) | 50 |
| Four big miners, from 2000 | 68 | 30 (44%) | 38 |
| One random miner (median run), from 2000 | 68 | 12 (18%) | 56 |
| GDX, from 2006 | 42 | 11 (26%) | 31 |
Read the first row with care. The four big miners are the survivors, chosen with 2026 hindsight, and most of their wins come from the late 1980s. From 2000 on, the same four lost to gold more often than they won. A typical random miner lost to gold in more than four out of five ten-year plans. GDX, the simple way to own the whole sector, did best at 35% more than gold and worst at 36% less, and lost three times out of four. With five-year plans the picture is the same: the four big miners won 64% of the time and GDX 35%.
Look at the red bars again and one decade does nearly all the damage. For the four big miners, gold won 41 of the 44 plans started from 1999 to 2009, and only 9 of the other 83. So is gold just a “lost decade” trade? Two things say no. First, it was not the stock market’s lost decade that hurt the miners. Plans started from 2004 to 2009 beat gold handily in the S&P 500, and still lost to gold in the miners. What hurt was buying miners during gold’s own bull run, when costs and acquisitions ate the higher gold price. Second, the “only then” holds for the four survivors, not for the sector. The XAU index, price only, beat gold in just 19 of 55 plans started from 1985 to 1998. And from 2010 on, GDX lost 17 of 28 plans and the median random miner 16 of 28, even in the miners’ good years.
Drawdowns: what holding miners actually feels like
CAGR hides the part that makes people sell. Gold’s worst fall in this sample was 47%, from December 1987 to July 1999 on monthly averages. That is already ugly: in July 1999 a quarterly gold buyer was 29% under the total money they had put in. Go back further and gold spent 26 years, from January 1980 to May 2006, below its 1980 average price.
The miners are in another category. Every one of the four big names fell 75% to 87% from a peak. The basket fell 76%. The quarterly account in the basket, even with new money cushioning it, fell 76% from its high, from December 2010 to September 2015. GDXJ fell 87%. These were not quick crashes. From September 2011 to September 2015, GDX holders watched 77% of the money disappear over four years while gold was down 39%.
If you want gold because it holds up when stocks do not, miners do not do that job. In the 2008 crash GDX fell 57% on monthly averages, against 49% for the S&P 500 total return index. Miners trade like small-cap stocks with a gold price attached.
What this test does not prove
The early years carry survivorship bias, and it is not small. Before 2000 the only miners with price data are the ones that still exist. Homestake, Placer Dome, Echo Bay, Battle Mountain, and Pegasus Gold are missing. The XAU check above shows how much that matters: from 1985 to 1999 the sector lost money while the four survivors doubled it. That is why the random draws start in 2000, and why the 1985 basket is hindsight and not a strategy.
After 2000 the pool does include failures. Banro went bankrupt, Primero was sold for pennies, Allied Nevada is in from its bankruptcy filing. It still misses some: Great Basin Gold, Jaguar Mining, Crystallex, and Allied Nevada before 2015 are not in my data. Takeovers like Placer Dome, Glamis, or Meridian mostly ended at a premium, so leaving them out matters less. The missing bankruptcies mean the random results after 2000 are still a little too kind to the miners.
The “big four” are hindsight. They are the four that were large enough in 1985 to have clean data and still matter today. A 1985 investor would not have known that Agnico Eagle would be the one to make 21 times buy-once.
No taxes, no trading costs, no storage or insurance on physical gold. Dividends are reinvested at the monthly average price with no withholding tax, which flatters the South African and Canadian ADRs a little for a US investor. And the end date is October 2026, during a gold price near all-time highs. If you read this after a 40% miner selloff, the miner rows will look a lot worse. That is the point of the rolling windows.
For context, the S&P 500 total return index beat both. The same $1,000 a quarter from January 1988 became $2.23 million in the index (11.1% a year on the money in) and $1.11 million in gold (8.5%). I wrote about why holding the broad index is hard to beat in time in the market vs timing the market, and what happens when you try to cherry-pick inside it in the odds of beating the S&P 500.
The honest verdict on gold vs gold miners
Over 41 years of quarterly buying, a handful of big miners ended slightly ahead of gold, and only because they paid dividends. On price alone, gold won. Valued just 22 months earlier, gold won outright in every window I ran. The sector ETF, GDX, lost to gold after 20 years even with the 2025 to 2026 boom in its favor. In ten-year windows, gold beat GDX almost three times out of four.
If you want gold exposure, own gold. A quarterly plan in the metal got you most of the return of the miners, with about half the drawdown and none of the company risk. If you buy miners, treat them as a leveraged, cyclical stock position. Use a small basket, not one name, and expect to sit through a 75% fall at some point. Buying them because gold “has to” go up gets the bet wrong: you also need the margins, the managers, and the end date to cooperate.
I keep the quarterly habit idea from my DCA vs market timing test: pick the instrument once, buy on schedule, and do not let the last two years choose it for you. That rule matters more with miners than with almost anything else.
FAQ: gold vs gold miners
Is it better to buy gold or gold mining stocks?
In this 1985 to 2026 test, a quarterly plan in four big miners ended at $1.51 million and gold at $1.24 million on the same $167,000. Without dividends the miners lost. Valued at December 2024, gold won by a wide margin. For gold exposure, gold itself was the steadier choice. Miners were a leveraged bet on how the gold cycle ends.
Did GDX beat gold?
Not quite. $1,000 a quarter from July 2006 became $242,572 in GDX and $250,273 in gold by October 2026. Buy-once, GDX returned 4.9% a year against 9.7% for gold, with a 77% drawdown. GDX beat a gold plan in 11 of 42 ten-year windows.
How much do dividends matter for gold miners vs gold?
They decide the 41-year result. Dividends added 46% to the four-miner plan from 1985. With them the basket ends at $1.51 million, without them at $1.04 million, behind the $1.24 million gold plan. Gold pays nothing, so miners need the income to keep up.
What are the odds a random gold miner beats gold?
From 2000 to October 2026, a single random miner bought every quarter beat the same plan in gold 52% of the time. Five random miners won 67% of the time. Valued at December 2024, one random miner beat gold in only 18% of runs. In the 68 ten-year windows starting from 2000, the median random miner lost to gold 56 times.
Does the test include gold miners that went bankrupt?
From 2000, yes, where prices exist. Banro, Primero, and Allied Nevada are in the random pool. If a held miner goes bankrupt or is bought, the last price is the exit and the money moves to another random miner. Before 2000 I have no prices for the companies that disappeared, so I used the XAU sector index as a check. It lost money from 1985 to 1999 while the four survivors doubled.
How deep do gold miner drawdowns get?
Very deep. Newmont, Barrick, Agnico Eagle, and Gold Fields each fell 75% to 87% from a peak at some point since 1985. GDX fell 77% and GDXJ 87%. Gold’s worst fall in the same sample was 47% on monthly averages.
Do gold miners protect a portfolio like gold does?
Not reliably. Miners are stocks with operating leverage, debt, and management risk. In the 2008 crash GDX fell 57% on monthly averages, the S&P 500 total return index 49%. From 2011 to 2015 GDX lost 77% while gold lost 39%.
Why did miners lag gold from 2001 to 2015?
Costs rose with the gold price, ore grades fell, and many miners spent heavily on acquisitions near the top. After 2011 the sector booked billions in write-downs. The Philadelphia Gold/Silver index went from $1.33 for every $1 in gold in 1996 to 16 cents in 2015, price only.
Which miners were in the random pool?
28 gold-first producers listed in the US, including names that disappeared: Goldcorp, Yamana, Newcrest, Kirkland Lake, Tahoe, Klondex, Richmont, AuRico, New Gold, Golden Star, Pretium, Primero, Timmins Gold, Banro, and Allied Nevada. Royalty and streaming companies, silver-first miners, and explorers are out. Draws start in 2000, because before that only the survivors have price data.
What data did this gold vs gold miners test use?
World Bank monthly average gold prices, and Yahoo and Tiingo daily dividend-adjusted prices averaged by month for miners and ETFs. $1,000 bought every January, April, July, and October from April 1985, valued at October 2026. No taxes, costs, or storage fees.



