Add a professional touch to your finances

“Don’t put all your eggs in one basket” – All of us must have heard this proverb, which roots for adding diversification to anyone’s investment portfolio. While the idea of portfolio diversification is a well-researched topic in the field of traditional finance (stocks, bonds etc.), application of diversification in cryptocurrency portfolios is rare. Through this article, we analyze whether we can truly diversify a portfolio of cryptocurrencies from both short-term trading & long-term investing perspective.

To carry out this study, we’re going to take coins which have a feasible underlying project and stand for practical application development (as you might’ve already guessed, we aren’t gonna include any meme coins like DOGE, SHIBA etc. in this analysis). Based on sufficient market capitalization & trading volumes, we’ve selected the following 10 coins:

  1. Bitcoin (BTC)
  2. Ethereum (ETH)
  3. Binance Coin (BNB)
  4. Cardano (ADA)
  5. Ripple (XRP)
  6. Polkadot (DOT)
  7. Uniswap (UNI)
  8. Chainlink (LINK)
  9. Litecoin (LTC)
  10. Polygon (MATIC)
Fig 1. Market capitalization of sample coins

As seen in Fig 1, BTC and ETH lead all other coins with massive market capitalization owing to their early launch & widespread adoption. The lowest market capitalization among all the coins in our sample belongs to MATIC (recently rebranded as Polygon), a Layer-2 Ethereum based platform of Indian origin. Now, we look at the trading volumes of these coins in the past week.

Fig 2. Trading volumes of sample coins

A similar trend is observed for trading volumes (Fig 2) with the exception of XRP & LINK which have considerably high Traded Volume/Market cap ratios compared to its peers. Inspite of a high market cap, Uniswap has faced relatively thin trading in the past week.

Investment & Diversification

For analyzing diversification in an investment portfolio, we’ve taken weekly data for the last 3 years (July, 2018 – July, 2021). An interesting principle behind diversification is the degree of correlation between components. Proper diversification is only possible among assets/components with low mutual correlation.

Fig 3. Pairwise correlation heatmap (Investment portfolio diversification)

Fig 3 displays the degree of correlation between the crypto pairs with BTC, ETH, ADA, XRP, LTC & BNB displaying strong correlation amongst each others. These coins have large market cap and significant project developments (forks, testnet & mainnet implementations etc.) which move the prices of all these coins together during certain periods. DOT, UNI & MATIC display relatively lower correlation with others. This implies that if an investor wants to create a long-term portfolio of coins, a combination of BTC, ADA, DOT, MATIC & UNI is likely to provide a better shield against unsystematic risk compared to an alternative portfolio of BTC, ETH, ADA, XRP & BNB.

Diversification in the Short run (Trading)

For testing trading portfolio diversification, we need access to high-frequency data. Firstly, we set up a JSON link between Power BI & Coinmarketcap API (learn how to do this here). Next, we create a charting algorithm using Python using an interval function [max(15,dP)]. This captures price movements after a pre-set interval of 15 sec or the first change in coin’s price (useful in case of illiquid coins). Since all the coins used in our study are relatively liquid, a 15 sec interval was maintained between 2 data points.

Fig 4. Implementation of the charting algorithm on ETH ticker

For trading portfolio diversification, we’ve taken data across 2 periods:

  1. Bull market (Mar – Apr, 2021)
  2. Bear market (Jun – Jul, 2021)

A similar heatmap (pairwise correlation) yields some interesting results.

Fig 5. Long-run & Short-run correlation heatmap comparison

As observed from Fig 5, short-run coin returns exhibit weak correlation with same-period peer returns. However, it is inconclusive of the direction of flow of information in crypto markets i.e. does BTC price change affect the performance of other coins or not? Traders often look at BTC/ETH prices to place their trades on other coins. We all know that cryptos are widely traded using Algos & Bots which use lagged return models to execute trades. So, the correct test would be price change of a coin at period ‘t’ with price change of BTC at period ‘t-1’ (lagged return).

Market Efficiency & Information Transmission

We divide our data into 3 broad categories:

  1. Normal Activity (NA) [Days on which overall crypto market cap change (absolute) was <3%]
  2. High Activity (HA) [Days on which overall crypto market cap change (absolute) was between 3% & 7%]
  3. Super Activity (SA) [Days on which overall crypto market cap change (absolute) was >7%]

We create a dataframe for all NA data during the bull market of Mar – Apr & name it NA1 and another dataframe for all NA data during the bear market of Jun – Jul & name it NA2. Similar approach is followed to obtain dataframes HA1 & HA2 and SA1 & SA2. Now we carry out separate zero-intercept RLS regressions for all coins in each dataframe (coin return at period ‘t’ vs BTC return at period ‘t-1’).

Fig 6. Inter-crypto sensitivity coefficients with lagged BTC returns

Fig 6 displays the most significant results in our entire analysis. The sensitivity co-efficient acts as an analogy to Beta (systematic risk indicator) used in case of stocks. The co-efficient between BTC & BTC(-1) returns indicate an Auto Regressive (AR) relationship. The key takeaways from Fig 6 are as follows:

  1. On normal activity days, BTC exhibits a mean-reverting tendency w.r.t its lagged return. Coins like ETH & BNB follow appx. unit sensitivity co-efficient w.r.t BTC(-1) returns, stabilizing the market from wide fluctuations & restricting overall m-cap change within +/- 3%.
  2. ADA, XRP, LINK & UNI have lower dependency on lagged BTC returns, creating a strong foundation for a well-diversified trading portfolio.
  3. MATIC has displayed high volatility across all sample periods and has a magnified effect of prior period BTC returns on its own price movements.
  4. In case of super activity days (huge rallies or crashes), a change in BTC price is likely to have a magnified effect on the returns of most of the other coins in our sample (directly proportional).
  5. A crypto trader can utilize the sensitivity co-efficient to take levered bi-directional trades (Buy DOT spot + Sell BTC futures on a SA1 day/Sell MATIC spot + Buy BTC futures) where allowed, to earn good hedged profits.

Food for Thought

Unlike traditional financial assets, cryptocurrencies don’t have any collateral to support their values. Any project needs months to get implemented and sometimes years to earn profits. Then, how do cryptos rise/fall by 50% in a week? How to select the correct crypto for trading/investing? Stay tuned for the answers to all these questions & many others you might have regarding the extremely interesting but dangerous world of cryptocurrencies.

Sayantan Ghosh

MBA (Finance), FMVA®