Are Cryptocurrencies “Diverse”?
Portfolio Construction for Hodlers
Ray Dalio, the founder of Bridgewater Associates, recently made the following comment on cryptocurrencies:
“…You want a properly diversified portfolio… Question is…
Do you have enough diversity of these alternative currencies?”
This is a big question and merits a lot of research. I have limited space in this short article, but we can take a first step. If you are pressed for time you can skim to the TL;DR at the end of the article.
The Big Four
To begin with, let’s look at daily returns for Bitcoin (BTC), Ether (ETH), Litecoin (LTC) and Ripple (XRP), aka “the Big Four”. If you haven’t heard them referred to in that fashion before it’s ok because I just made it up.
The average correlation of the daily returns of these four coins from February 2016 through May 2021 is approximately 0.38. This is reasonably low and suggests that investing in these four coins offers some diversification benefit.
Unfortunately, the long-term average correlation does not tell the whole story. From early 2016 through the end of 2019 the average correlation was 0.185, from 2019 through May of 2021 that number has risen to 0.744.
We can see this by looking at the rolling 30-day average correlation (I’d use a longer window for traditional assets, but 30 days seems informative for cryptocurrencies due to a. their high volatility and b. the rapid pace of innovations in the space).
Coindesk 20 (-2)
Four of anything is unlikely to offer a great deal of diversification potential, so let’s look at a broader set of currencies. The CoinDesk 20 Index seeks to be representative of the cryptocurrency opportunity set. For the next few pieces of analysis, I will use daily returns for the current constituents, less the two stablecoins (USDT and USDC).
The average correlation for the 18 currencies is 0.594 for daily returns from the end of December 2020 through late May 2021 — about 150 observations as cryptos trade 7 days a week.
Let’s dig a little deeper:
The barplot below shows the average correlation of each coin to all of the others in our sample. Filecoin, Ripple and Uniswap have the lowest average correlations with the rest of our sample, while Tezos, Chainlink and Litecoin have the highest average correlations with the sample. Bitcoin is in the middle of the pack, which I found a bit surprising.
On a pairwise basis, we can see that Filecoin, Ripple and Uniswap show up in most of the 10 least correlated pairs, and Chainlink, Tezos and Litecoin show up in most of the 10 most correlated pairs. This kind of information can be helpful if we are looking to diversify a core allocation to a particular currency or a concentrated sector bet.
I like to think of myself as a “good” investor, but I do not want to fool myself into thinking that I am much better than I really am. One of the tools I use to guard against fooling myself is to to look at the distribution of potential outcomes for parameters that I can change, for example, selecting investments or their portfolio weights. It can give a better picture of what is realistic and where the vicissitudes of the markets may take us.
This is a very simple simulation where we pick five coins out of a hat (without replacement), store the average correlation, and repeat thousands of times.
The table below shows the summary data for the average correlations of 10,000 randomly selected portfolios of 5 currencies from our sample. The chart is a histogram of the same data. Picking 5 coins out of a hat yields average correlations that range from 0.43 to 0.78 with an average of just under 0.60 (the dark blue line on the histogram).
Ray’s Pot of Gold
So, what does Ray Dalio consider to be enough diversification? Ray has often spoken of the “holy grail of investing”, which he characterizes as finding 15 or so uncorrelated return streams. If they are truly uncorrelated, they don’t have to be great or even good, they just need to be ok to generate great risk-adjusted returns in combination. This sounds easy, but in fact it is very hard.
Assuming the volatilities and correlations are stable; a portfolio of 5 assets that have a 0.60 correlation to each other (the mean from the simulation above) will only reduce the volatility by about 15%. At a correlation of 0.40 (the left tail of the distribution above), the volatility will be reduced by about 25%. For either scenario most of the benefit comes from the first five assets and there is not much improvement to be had from adding more.
Based on this small sample, it seems we can hope for a modest amount of risk reduction for a concentrated portfolio of cryptocurrencies if we are careful (or lucky) about picking the constituents.
While the correlation of our “big four” sample has increased in recent years, the proliferation of cryptocurrencies with differing purposes appears to be driving an increase in diversification opportunity for investors. The current opportunity set for creating a diversified portfolio is modest, but I believe that as the asset class continues to evolve and mature, we will see an even greater potential for diversification.
It should be noted that I used simple average correlation as a heuristic for diversification in this article. There are many other ways to think about and measure dependence between assets, which I may explore in the future.
Thank you for reading.
- Correlation between the oldest, largest cryptocurrencies has increased markedly.
- A quick analysis of the current constituents of the Coindesk 20 Index indicates the potential for some diversification.
- As the market matures I expect diversification opportunities to increase.
References & Resources
- Sample R Code: https://github.com/rufusrankin/jff-are-cryptos-diverse/blob/main/code
- Ray Dalio quotation source: https://www.zerohedge.com/markets/i-have-some-bitcoin-dalio-prefers-bitcoin-bonds-thinks-ethereum-more-efficient
- Ray Dalio breaks down his “Holy Grail”: https://www.youtube.com/watch?v=Nu4lHaSh7D4
- Cryptocurrency price data from riingo/Tiingo
This article is for information purposes only and does not constitute investment advice. Any opinions are those of the author and do not represent those of Ampersand, Drexel University or any of their affiliates.
Join Coinmonks Telegram group and learn about crypto trading and investing