Diversification can reduce risk

Even though cryptocurrencies are (one of the) highest-risk investments you can make, it can still pay off to properly diversify a cryptocurrency portfolio to reduce some of the inherent risk.

Risk types

Portfolio risk always consists of two parts: systematic risk and specific risk. Systematic risk includes, among others, the uncertainty regarding the future use of cryptocurrencies as a payment system. It is the type of risk that affects the market as a whole. The question whether, for example, Bitcoin will remain a part of this could be indicated as specific risk to this cryptocurrency. This specific risk is considered diversifiable. By spreading an investment accross multiple cryptocurrencies, it is possible to reduce specific risk until only systematic risk remains while optimizing returns.

Reducing risk

To get an idea of how this works, it is required to know the riskiness, or volatility, of the concerned cryptocurrencies as well as the correlation coefficients between cryptocurrency pairs. The correlation coefficient represents a measure of strenght of the relationship between two cryptocurrencies. This information can be derived from historical data. For the following example hourly observations from Dogelytics are used. The observed correlation coefficients and volatilities are shown below.

Correlations and Volatilities

It can be seen that the lowest correlation is between Dogecoin and Peercoin. A correlation coefficient can take a value between +1 and -1, where +1 indicates a perfect linear relationship (obviously coins are perfectly correlated with themselves) and 0 indicates no linear relationship. -1 would indicate a perfect negative linear relationship. The 0.524 for DOGE/PPC signals a moderate positive linear relationship. Assume a portfolio of 20 percent invested in Dogecoin and 80 percent invested in Peercoin. If these coins were perfectly calculated, one could simply add up 20 percent of 0.86 percent (DOGE hourly volatility) and 80 percent of 0.60 percent (PPC hourly volatility) resulting in 0.65 percent. Taking into account the correlation coefficient of 0.524, this would change to a little below 0.59 percent. The latter is less risky than an investment in Peercoin alone, and comes down to a relative reduction of volatility by nearly two percent. If 10 percent would be invested in Litecoin at the expense of Peercoin, the hourly portfolio volatility would go down further to less than 0.58 percent. A relative reduction of four percent.

Perhaps it does not seem like a lot, but this is only a rough calculation on hourly data and just three cryptocurrencies are combined. Most importantly, it should be noted risk was reduced by investing in cryptocurrencies that are more of a risky investment by themselves. An optimal portfolio would combine all cryptocurrencies, and there are already over one hundred cryptocurrencies available on Cryptsy. This also means that there will be over 5000 cryptocurrency pairs. Many of these pairs will be less or even negatively correlated. The average investor is not likely to include all of these, but it offers many possibilities to spread out an investment and significantly reduce any specific risk. A portfolio of just ten different cryptocurrencies should be enough to capture most of the diversification benefit.

  • Nice suggestions on portfolio risk reduction! If you’re looking to expand the analysis to daily log return relationships, I’ve been tracking this correlation data at http://coincorrelation.com .
    Hope it helps!

    • Nice tool. I provide some basic data on http://digilytics.net, correlation was scheduled to be included although it would not be this extensive. Do you have the statistical significance of the correlations? I could not find it, but this would make it better suited to actually use it for trading.

      • Thanks for the tip – I’ll include statistical significance on the list of future features. I was also going to add a correlation stability metric for the shorter lookback periods, based on the mean and standard deviation of the correlation values over time, and use that as a confidence indicator.