Bitcoin vs S&P 500 and Gold: Correlation Measured

Across every US trading day from October 2016 to October 21, 2026, the correlation between bitcoin's daily return and the S&P 500's was 0.27, and with gold it was 0.11. Those two numbers hide the fact that the same pair has been negative in some years and above 0.5 in others, so a single correlation figure always describes one window, one return definition and one way of lining up the dates, and all three are stated below.

A reading of 0.27 means about 7 percent of the day-to-day variation in bitcoin's return lines up linearly with the S&P 500's (0.27 squared is 0.073). The rest is bitcoin's own. That is a different story from the one told by the charts in which both lines climb together, and the first section shows where that picture comes from.

The 0.91 that means nothing

Plot bitcoin's price against the S&P 500's index level over the same 2,512 trading days and the correlation comes out at 0.91. Against a gold fund's share price it is 0.82. Both numbers are produced by the same mistake: two series that trend upward over a decade correlate highly whether or not they have anything to do with each other, because each level is mostly the sum of everything that came before it. A correlation of levels measures shared trend, not shared behavior.

The standard fix is to correlate returns, the percentage change from one observation to the next. A daily return has no memory of the level, so the trend drops out and what remains is whether the two assets moved the same way on the same day. Every figure below from here on is a correlation of returns.

How the series were built

Four choices decide what the numbers mean, and a different choice in any of them gives a different answer.

The data. Bitcoin is the Coin Metrics PriceUSD daily series from the community API, pulled October 23, 2026. The S&P 500 and the Nasdaq Composite are index closes from the Yahoo Finance chart interface, and gold is the closing price of the SPDR Gold Trust ETF (GLD) from the same interface, so gold here means the price of a fund holding bullion and not the London spot fixing. The S&P 500 closes were checked against the Federal Reserve Bank of St. Louis FRED series SP500, which carries S&P Dow Jones Indices data: all 15 of the latest FRED observations (October 1 to October 21, 2026) and 11 earlier dates, from January 3, 2017 to June 30, 2026, matched to the cent. The Nasdaq and GLD closes could not be checked against an official print and are labelled secondary. The sample ends on October 21, 2026 because that is the last day FRED had published when this page was written.

The calendar. Bitcoin trades every day and US markets do not. Each return here is measured between consecutive US trading days, using bitcoin's price on those same two dates, so a Monday return for bitcoin spans the whole weekend, exactly like the S&P 500's Friday-to-Monday return does. Of the 2,511 return observations, 548 (21.8 percent) cover more than one calendar day. Comparing the S&P 500's Monday move with only bitcoin's Sunday-to-Monday move would pair a three-day change with a one-day change.

The clock. Coin Metrics describes PriceUSD as "the fixed closing price of the asset as of 00:00 UTC the following day (i.e., midnight UTC of the current day)." A row dated October 21 is therefore bitcoin's price at the end of October 21 UTC. The US cash session closes at 4 p.m. New York time, which is 20:00 UTC in daylight time and 21:00 UTC in standard time, so bitcoin's observation falls 3 to 4 hours after the stock close on the same date. News that arrives in that gap moves bitcoin's number but not that day's stock close, a small mismatch that can only lower the measured correlation.

The return. Daily close-to-close percentage change. A weekly version, using the last trading day of each week, is shown as a check.

Five days by hand

The calculation is short enough to do on paper, and doing it once shows why short windows mislead. Take the five most recent trading days in the sample. The percentage returns, bitcoin first and the S&P 500 second, were:

The mean return is 0.652 percent for bitcoin and 0.101 percent for the S&P 500. Subtract each mean from each day to get deviations: for bitcoin 0.106, -1.355, -0.186, 0.546 and 0.889; for the S&P 500 0.225, -1.193, 0.308, 0.183 and 0.478. Multiply the two deviations day by day and add them up: 0.024 + 1.618 - 0.057 + 0.100 + 0.425 = 2.109. Square each set of deviations and add them: 2.971 for bitcoin and 1.831 for the S&P 500. The correlation is 2.109 divided by the square root of 2.971 times 1.831, which is 2.109 divided by 2.332, or 0.90.

A correlation of 0.90 over those five days and 0.27 over 2,511 days describe the same two assets. The five-day figure is dominated by a single day, October 16, when both fell and each deviation was large, which contributed 1.618 of the 2.109 total. With only five points one pair can set the answer, which is why very short windows swing so widely.

The full-sample numbers

The last row is a calibration. Two US stock indices that overlap heavily correlate at 0.95 on the same method, so 0.27 is far from what two tightly linked assets look like. Over the sample bitcoin's daily standard deviation was 4.17 percent, against 1.14 percent for the S&P 500, 1.40 percent for the Nasdaq Composite and 1.04 percent for GLD. A regression of bitcoin's return on the S&P 500's gives a slope of 0.98, meaning that on average bitcoin moved about one percentage point per point of S&P 500 movement, with a great deal of scatter around that line: the correlation squared, 7.3 percent, is the share of variance the line explains.

Direction agreed more often than size did. Bitcoin and the S&P 500 moved in the same direction on 60.1 percent of the 2,511 days.

Year by year

The sequence has a shape that a full-sample number conceals. For 2017 to 2019 bitcoin's link to the S&P 500 was close to zero or negative. It rose in 2020, was highest in 2022 at 0.57, fell back to 0.16 in 2023, and has been between 0.35 and 0.46 in each year since. Gold's link to bitcoin stayed small in most years but reached 0.30 in 2020 and 0.31 in 2026, a year in which the S&P 500 and gold also correlated at 0.40, the highest in the table. In 2026 all three pairs are positive at once, which this method describes but cannot explain.

Weekly figures are lower than daily ones in most years. The sharpest gap is 2024, 0.35 daily against 0.13 weekly, which says that year's daily relationship came from a set of short-lived same-day moves that did not carry over across the week.

How much the window changes the answer

A rolling correlation recomputes the figure over a moving window. The window length is the largest single source of disagreement among published numbers, because short windows are noisy. The table runs 30 and 90 trading-day windows across the whole sample.

The medians are taken over all windows of each type. The short-window bitcoin and S&P 500 series spent a fifth of its history above 0.5 and a fifth below zero, which is the same asset pair and the same data. On the latest windows ending October 21, 2026, bitcoin and the S&P 500 read 0.58 over 30 trading days and 0.38 over 90; bitcoin and gold read 0.23 over 30 and 0.53 over 90. Neighboring windows share most of their days, so the shares in the last two columns describe this one history and are not probabilities.

The spread inside a single year is as informative as the yearly average. The table gives the range of the 90-trading-day bitcoin and S&P 500 correlation across windows that end within each calendar year.

The medians are taken over the windows listed in each row. In 2023 the 90-day figure ran from 0.05 to 0.60 within twelve months, and in 2020 it ran from minus 0.04 to 0.60. Every window ending in 2022 and 2026 stayed above 0.3, and every window ending in 2019 stayed below 0.13.

When the S&P 500 had its biggest days

Correlation is an average over every day, but the days that matter to most readers are the extreme ones. These are the six largest S&P 500 falls and rises in the sample, with bitcoin's and gold's returns over the same trading-day interval.

Ten of these twelve days are in 2020, nine of them in March and April, which is why a single month can move a correlation. The table's clearest counterexample is April 4, 2025, when bitcoin rose 1.02 percent while stocks fell almost 6 percent. The worst day, March 16, 2020, followed a weekend, so bitcoin's figure on that row covers the same three calendar days as the stock move.

Splitting all 2,511 days by the size of the S&P 500's move shows where the correlation lives. On the 610 days when the S&P 500 moved by more than 1 percent in either direction (24.3 percent of days), the correlation with bitcoin was 0.46. On the other 1,901 days it was 0.09. Among the 276 days when the S&P 500 fell by more than 1 percent, bitcoin's mean return was minus 1.78 percent and its median minus 1.45 percent; among the 334 days when it rose by more than 1 percent, bitcoin's mean was plus 1.73 percent and its median plus 1.26 percent. So in this sample bitcoin tended to follow big stock days in the same direction, and ignored small ones.

The response was also lopsided. Fitting a line of bitcoin's return against the S&P 500's separately on the 1,137 days when the S&P 500 fell and the 1,373 days when it rose gives a slope of 1.26 on down days and 0.75 on up days. Put plainly, in this sample each point of S&P 500 decline came with about 1.26 points of bitcoin decline on average, while each point of S&P 500 gain came with about 0.75 points of bitcoin gain. One day in the 2,511 had an exactly flat S&P 500 close and is in neither group. The slopes carry the same scatter as the correlation does, and they describe the past.

The alignment trap

The calendar and clock choices above are not small print. Shift bitcoin's series by a single day relative to the stock series and the full-sample correlation with the S&P 500 collapses.

A mismatch of one day wipes out almost all of the relationship, which has two consequences. A careless comparison, for example one that pairs bitcoin's price at one fixed hour of a New York calendar day with stock closes, can understate the link or overstate it depending on where the mismatch lands. And in this sample there is no sign that bitcoin's move on one day lines up with the stock market's move on the next, in either order; the link is contemporaneous.

Multi-day gaps matter in another way. The 548 observations that span a weekend or holiday correlate at 0.20, against 0.30 for the 1,963 single-day observations. Bitcoin trades through weekends when stock markets are closed, so a Monday return for bitcoin combines the days on which it could react to news by itself with the day when the stock market could react as well.

Reconciling published numbers

A web search on October 23, 2026 for bitcoin and S&P 500 correlation returned several 2026 readings that disagree with one another. The result summaries (the pages themselves were not opened) said that one aggregator table listed a 30-day figure of 0.74, that a German financial news site reported a 30-day rolling correlation at a multi-year high of 0.81 in May 2026, and that a third data site showed a near-zero reading as of April 24, 2026. A fourth source, a personal post, reported 0.44 for the 2026 year to date through August 21. None of these gives enough detail about the return definition, window or data to repeat, so none is used as a figure on this page.

The only one that can be checked is the last. Using this page's method, the daily bitcoin and S&P 500 correlation from January 1 to August 21, 2026 is 0.44 on 160 trading days, which matches. For the 30-day readings, this method gives 0.47 for the window ending April 24, 2026, 0.47 for May 29 and 0.48 for June 5, with the highest 30-trading-day figure for any window ending in May at 0.49 on May 27. No 30-trading-day window ending in 2026 reached 0.74. One possible cause is the window: a 30-calendar-day window contains only about 21 trading days, and a calculator that pairs bitcoin's seven-day week with a stock series that skips weekends makes different pairs from the ones used here. This page cannot tell which explanation applies to any given source; the point is that two readers who both say "30-day correlation" may be computing different things.

What 0.27 cannot tell you

A correlation coefficient is sensitive to a handful of extreme observations, and this one is. March 2020 contributes 22 trading days in which the correlation between the two was 0.63. Removing those 22 days lowers the full-sample daily bitcoin and S&P 500 correlation from 0.27 to 0.23 on the remaining 2,489 days, and lowers 2020's figure from 0.44 to 0.24 on 231 days. The relationship does not disappear without the crash, but a fair share of what looks like a durable link in the 2020 row is one month.

It is also a statement about linear co-movement in returns and nothing else. It does not say bitcoin is caused to move by stocks, or the reverse. It does not say whether bitcoin will fall when stocks do next time: a correlation of 0.46 on big stock days in the past leaves wide room for the day on which bitcoin goes the other way, as it did on April 4, 2025. And it says nothing about whether bitcoin is a hedge, a risk asset or a diversifier, because those depend on how much of a portfolio is involved, over what horizon, and against what other holdings.

The gold figures carry an extra caution. GLD is a fund that trades during US hours, so its close is aligned with stocks, but gold itself trades around the clock in London and Asia, and a spot-gold series would pair differently with bitcoin's midnight-UTC fixing. For the price level of bitcoin itself, see the bitcoin market page; for how far bitcoin's own daily moves reach, the volatility history and the bear market measurements give the drawdown and volatility figures on a stated basis. How interest-rate news fits into bitcoin's daily moves is covered in the page on Fed rate decisions.

Frequently Asked Questions

What is the correlation between bitcoin and the S&P 500?

On daily returns between US trading days from October 25, 2016 to October 21, 2026 it is 0.27, and on weekly returns it is 0.19. The yearly figure has ranged from minus 0.12 in 2019 to 0.57 in 2022 and was 0.46 in 2026 to date, so no single number describes it. The method is on this page: Coin Metrics bitcoin prices, S&P 500 closes checked against FRED, same-date alignment.

Is bitcoin correlated with gold?

Weakly. The full-sample daily correlation with the GLD gold fund is 0.11. It has been near zero in most years and was 0.30 in 2020 and 0.31 in 2026. Rolling 90-trading-day readings ranged from minus 0.28 to 0.59, and the latest reading ending October 21, 2026 was 0.53.

Why do different websites show different correlations for the same two assets?

Because they differ in return definition, window length, how weekends are handled and where the dates are aligned. A correlation of price levels gives 0.91 for bitcoin and the S&P 500 in this sample, against 0.27 for daily returns. A one-day misalignment turns 0.27 into about zero. Pages rarely state these choices.

Does bitcoin go up when the stock market goes up?

On days when the S&P 500 rose more than 1 percent, bitcoin's mean return was plus 1.73 percent and its median plus 1.26 percent, across 334 days. On days when stocks fell more than 1 percent, its mean was minus 1.78 percent across 276 days. On days with small stock moves the relationship was weak, at 0.09. These are historical averages and individual days differ.

Is a high correlation a reason to expect bitcoin to fall when stocks fall?

This page does not support that inference. It describes what happened in a past sample, including a 2025 day on which bitcoin rose 1.02 percent as the S&P 500 fell 5.97 percent. Correlations change from year to year and window to window, and they do not describe causes or future moves.

How can I check these numbers?

Download bitcoin's daily PriceUSD from the Coin Metrics community API, the SP500 series from FRED, and a gold series of your choice. Keep only dates when US markets were open, compute the percentage change between consecutive such dates for each series using bitcoin's price on the same dates, and take the Pearson correlation of the two columns. Changing any step will change the answer.

Sources

CryptosEyes publishes general educational research, not investment, legal, or tax advice. This article reports statistics computed on historical prices with the method stated. A correlation describes the past, does not imply causation, and is not a forecast or a recommendation about any asset or portfolio.