We discussed in a previous article how correlation wasn’t necessary the right metric to look at when it comes to managing risk, and that beta (a function of correlation and volatility) is a much better risk measure when it comes to constructing a crypto portfolio . However, much of the industry still looks at correlation as the go-to metric and given the industry belief that cryptos are super correlated, we wanted to run the numbers ourselves to better understand just how correlated crypto assets really are. No one ever doubts the ability to construct a diversified stock portfolio — we wanted to take the first step here to address the question, “is it possible to construct a diversified crypto portfolio?”
Tldr – Cryptos are more correlated to each other than stocks during super volatile periods but have a similar correlation profile to stocks in normal market environments.
To run this analysis, we pulled historical year-over-year returns for every company in the S&P 500 and every crypto in the top 100 list going back to 2018 (from the beginning of the last crypto bear market). S&P 500 is generally viewed as the benchmark for US equities, and the top 100 list of cryptos makes up over 95% of the entire crypto market cap. Given this dynamic, we felt it probably made sense to use these two universes of assets.
Next, we took every stock in the S&P 500 and correlated it to every other stock in the S&P 500, using a one year rolling correlation and removing any instance of double counting along the way. We used a Pearson correlation here, as that’s the basic form of correlation that most folks have in mind. We then followed up this step by taking the average of all the correlation pairs (basically 125,000 pairs) to understand what the “average correlation” was for the S&P 500 index.
We ended by doing this with the top 100 cryptos (excluding stablecoin of course)…and were pretty surprised with the results:
The average correlation for cryptos fluctuated quite range, ranging from as high as almost 0.9 to as low as under 0.3. The average correlation for stocks experienced fluctuations as well, though not as wide as that of cryptos. Stocks reached a high of almost 0.6 and a low of under 0.2.
What is very interest here though is that the average correlation of cryptos actually trended down during the bear market from 2018 to 2020, and again after the initial surge in 2020 (reflecting the start of the COVID pandemic, keeping in mind that these are year over year numbers).
This shows that during very risk on and risk off events, correlations in crypto markets will spike and trend towards 1. However, this is true of almost all risk assets. In times of volatility risk market correlations, like that of cryptocurrencies, tend to go to 1. We’ve just used US equities, high yield, commodities, and cryptos below as an illustrative example. Similar risk-on baskets can be constructed including emerging market debt, emerging market equities, tech stocks, etc. — the conclusion remains the same regardless.
So compared to equity markets, what we’ve seen so far is that crypto markets tend to be more correlated in risk on or risk off events, but actually exhibit similar intra-asset class correlation outside of those periods of very high volatility where broader risk assets will trend towards a correlation of 1 anyways.
However, there’s one other thing that we should look at as well, and that’s how often are the correlation pairs in the crypto space actually positive? In other words, how often will one cryptocurrency trend whilst positive another is trending negative?
Here, you can clearly see that most cryptos will exhibit positive correlations most of the time (ie, when one goes up, the other goes up and when one goes down, the other goes down). In fact, this is most likely the reason that people say that crypto markets are extremely correlated. It’s that irrespective of the magnitude of the directional move or how high/low the correlations actually are, good days generally means a sea of green and bad days generally means a sea of red. It is very interesting to point out here as well that the percentage of positive correlation pairs also follows the average correlations — meaning positive correlations spike during periods of high market volatility and drop during normal market environments. And in fact, there are points where the percent of positive correlations match that of the S&P 500.
From the data we’ve seen here (and of course, given the nascent nature of cryptocurrencies, we must be mindful of time period bias and that previous performance patterns will likely change in the future), we can determine that whilst it’s definitely true that crypto markets can be highly correlated, this is not always the case. Crypto markets are highly correlated during severe risk on / risk off events. However, in this past cycle, their correlation profiles have actually reflected that of US equities in normal market conditions. If one were to have a portfolio of crypto assets and cash, what this means is that looking at the correlation metric, it’s feasible to create a crypto portfolio that is as diversified as an equity portfolio in normal market conditions. However, in risk off events, investors looking to construct a crypto portfolio will no doubt have to rely on equity a lot more than their investor counterparts to dampen portfolio volatility.
New to trading? Try crypto trading bots or copy trading