“As institutional investors [and retail traders] evaluate crypto assets, how can they think about properly assessing their risks, especially in the context of a broader, multi-asset class portfolio?”
The Pearson Correlation Coefficient quantifies the estimated strength of the linear association between two variables. It ranges from +1 to -1: +1 indicates a perfect positive linear correlation, -1 a perfect negative linear correlation, 0 indicates no linear correlation. The charts below represents the change in the correlation over time (rolling correlation with a rolling window width of 20 data points).
The Confidence Interval represents the interval in which the true Pearson correlation coefficient will be located with 95% probability. If the confidence interval includes the value of zero (it crosses the dashed line), the correlation can be regarded as non-significant (not different from 0).
As like this TwoSigma article: Risk Analysis of Crypto Assets - Two Sigma
to compare correlations, we first need to establish a universe of crypto assets to create correlation matrices. I have chosen for this universe of crypto assets BTC, ETH, the top 5 “data” coins as well as Doge and the S&P 500 for further analysis. It was found in the TwoSigma article above that Doge appeared most unique compared to their chosen universe so will the same follow here?
With the past 24hr spike in $NMR how do our correlations look?
As you would expect, $NMR was one of the most decorrelated assets over the past 24hrs from all other assets in our universe. BTC/ETH kept a very tight correlation of 98% while $NMR to $BTC and $ETH were 40% and 46% respectively.
How does this change over 30days and 1 year?
Over 30 days we see a big change from the 24hr window. NMR is mostly correlated with all assets in our universe, even with the S&P. $NMR keeps a correlation over 90% for both BTC and ETH. Interestingly though, Doge is the odd one out with less than 50% correlation to almost every asset.
Would this correlation hold out to 1 year?
This is where things get a bit interesting. $NMR is keeping a fairly tight correlation to most of the other “data” coins as well as Doge but when we look at correlation to BTC, ETH and the S&P we see a noticeable decorrelation. $NMR is very decorrelated from the S&P with a -29% correlation. Looking at BTC and ETH we see correlations of 51% and 32% respectively. Doge keeps its low correlation to BTC, ETH and the S&P 500 but interestingly, has a high correlation to mostly all “data coins”(Is NMR a meme coin?) . As with the TwoSigma article, BTC and ETH has kept a very consistent correlation to each other over various time frames.
In conclusion, although we find that BTC and ETH are highly correlated with each other, NMR over the long term does seem to show a deviation from BTC and ETH. It is also negatively correlated with the S&P.