Correlation in a Crash: Six Stocks Went From 0.38 to 0.90

Between 19 February and 23 March 2020, SPY (the S&P 500 ETF) fell 33.7% in 23 trading days. Six large stocks (AMD, AMZN, INTC, MSFT, NVDA and TSM, which is Taiwan Semiconductor Manufacturing) moved almost in step. Correlation measures how closely two prices move together. The mean correlation between each pair of the six was 0.90, against 0.38 on all days outside four large falls since January 1999.

This guide defines calm and stress, then measures all four falls in SPY of 20% or more since 1993. It ends with TLT (the long-term US Treasury bond ETF), GLD (the gold ETF) and Bitcoin, to see which kept their distance from SPY.

A return is the percentage change in a price over one trading day. Correlation compares two lists of daily returns. It is +1 when the two rise and fall together in exact proportion. It is 0 when there is no steady link, while -1 means one rises as the other falls.

Diversification means holding investments that do not move together, so that one fall is partly offset by another's gain. It relies on correlation staying low. The 60/40 portfolio is the classic mix.

For a group of investments, we use the mean pairwise correlation. A pair is two investments. Six stocks make 15 pairs. We calculate the correlation of each pair, then take the mean of the 15 numbers.

All returns here count dividends.

Stress is a fall in SPY of 20% or more from its highest close to that date. We measure from the high to the lowest close before SPY set a new high. We searched SPY's whole file, 8,493 trading days of prices from 29 January 1993 to 26 October 2026. It holds four such falls.

Three other falls came close: 19.0% in 1998, 19.3% in 2018 and 18.8% in 2025. They miss the 20% line. Their days count as calm, except 1998, which precedes our sample.

Calm is every other trading day from 25 January 1999. That date is the first return for all six stocks.

The table gives the four windows.

Mean Pairwise Correlation Of Six Stocks, Calm Days And Four Falls In SPY

Six stocks: AMD, AMZN, INTC, MSFT, NVDA and TSM. Correlation of daily returns, with dividends counted. Trading days count closes after the high close, up to the low close. Source: YX Insights price data

Chart 1 compares the calm figure of 0.38 with each fall. Correlation rose in every one: 0.47, 0.56, 0.90 and 0.69.

Of the 60 pair comparisons (15 pairs in each of four falls), 59 rose. The exception is one pair in the dot-com fall. In 2020, all 15 pairs sat between 0.80 and 0.94.

The dot-com rise was the smallest. To see how unusual each fall was, we took every run of consecutive calm days of the same length and calculated its mean pairwise correlation.

In the 2007 to 2009 fall, 2020 and 2022, none of the 4,060, 5,662 and 4,802 calm runs reached the stress figure. In the dot-com fall, 737 of 3,118 did, so that rise was not unusual.

The strongest calm run for 2020 covers 24 March to 24 April 2025, with a mean of 0.88. The 18.8% fall in early 2025 sits just under our 20% line.

Volatility is the size of daily price swings. We measure it as the standard deviation of SPY's daily returns, scaled to a year with the square root of 252 trading days. Our explainer on standard deviation in investing defines it.

SPY's volatility was 15.7% on calm days, against 23.6% in the dot-com fall, 38.1% in 2007 to 2009, 76.6% in 2020 and 24.2% in 2022.

The dot-com fall and 2022 had almost the same volatility, 23.6% and 24.2%. Their mean pairwise correlations were 0.47 and 0.69. Volatility alone did not set the correlation in these four falls.

We split the six stocks' daily moves into one shared pattern and the rest. The shared pattern explained 49% of their daily variation on calm days and 91% in 2020. The test gives the size of the shared move, with no view on its cause.

Chart 2 shows the mean pairwise correlation of the six stocks over the previous 63 trading days, about three months. It has 6,920 values, from 23 April 1999.

The median of the line is 0.43. The low was 0.10 on 25 July 2014. The high was 0.84 on 21 May 2020, in a window that still held March. On 26 October 2026 the line stood at 0.45.

Chart 3 shows the correlation with SPY of TLT, GLD and Bitcoin (BTCUSD), calm days against the falls where each has a full price history. Calm days start when each file starts: 2002 for TLT, 2004 for GLD and 2014 for Bitcoin.

We use SPY's trading days for all three, so that Bitcoin's weekend moves fall into Monday's return. Bitcoin trades every day of the year, while stocks trade about 252.

TLT had a correlation with SPY of -0.28 on calm days, -0.47 in 2007 to 2009 and -0.50 in 2020. In those windows TLT rose 25.0% and 14.2%, with SPY down 55.2% and 33.7%.

In 2022 the correlation was +0.03. TLT fell 29.3% while SPY fell 24.5%. Our explainer on the 2022 rate hikes covers that year.

GLD had a correlation of +0.09 on calm days, then 0.00, -0.01 and +0.09 in the three falls. Its return in the windows was +23.9%, -3.6% and -7.3%. Our guide to gold as a safe haven covers the longer record. Our explainer on how ETFs perform in a crisis covers more funds.

Bitcoin had a correlation of +0.16 on calm days, against +0.64 in 2020 and +0.57 in 2022. It has no reading for the two earlier falls.

The test uses six stocks. A set that spanned more industries could have a different calm figure.

We repeated the comparison with every stock in our files that has a full history for the window. The sets had 8, 9 and 10 names for 2007 to 2009, 2020 and 2022. The mean pairwise correlation was 0.52, 0.90 and 0.64 in those falls, against 0.36, 0.38 and 0.38 on calm days for the same names.

We also narrowed calm to days when SPY sat within 5% of its closing high, outside the four windows. The six-stock figure fell to 0.30 on those 3,251 days.

The 20% line, the six names and the 63-day window are our choices. The test covers four falls, which is a small sample.

Our explainer on volatility covers the swings that came with these falls.

In four falls in SPY of 20% or more, the six stocks moved more closely together than on other days. The rise was largest in 2020. TLT and GLD stayed below +0.10 with SPY in every fall tested, while Bitcoin reached +0.64 and +0.57.

The Academy is free to read, here on the web. The Long Trend, our free weekly letter, comes by email.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

Yes, among the six stocks we tested. Their mean pairwise correlation was 0.38 on calm days and 0.47, 0.56, 0.90 and 0.69 in four falls of 20% or more in SPY. Of 60 pair comparisons, 59 rose. Six stocks are a small sample, so the result does not describe every asset.

Our test shows that they do, but not why. One shared pattern explained 49% of the six stocks' daily variation on calm days and 91% in the 2020 fall. Volatility rose too, but the dot-com fall and 2022 had similar volatility and different correlations.

It depended on the asset and the fall. TLT had a correlation with SPY of -0.47 in 2007 to 2009 and -0.50 in 2020, but +0.03 in 2022. GLD stayed between -0.01 and +0.09 in all three falls. Bitcoin's correlation was +0.64 in 2020. Our test covers three falls and three assets.

We measure it as the mean of the pairwise correlations of daily returns inside the fall. We defined the fall as a drop of 20% or more in SPY, from its highest close to the lowest close before a new high. We compared it with calm days, which are the days outside the four falls.

DISCLAIMER: This article is intended for educational purposes only. Any information or analysis in this note does not constitute an offer to sell or a solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice, nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.