April 25, 2022
If you remember my April Fools Day post from a few weeks ago, I poked fun at the proliferation of new crypto coins. Most of them are scams. But what about the mainstream crypto coins, like Bitcoin, Ethereum, etc.? Are they a good investment? What’s not to like about a 100%+ annualized return in some of the crypto coins between their inception and their 2021 peak?
Well, those returns are “water under the bridge”. What matters to me today is the outlook for the crypto world going forward. In today’s post, I like to go through some of the reasons why I believe going forward, crypto looks like a sub-par investment. I currently don’t invest in crypto and I don’t think that anything more than a few % of the portfolio seems prudent. Let’s take a look…
Just to be sure, cryptocurrencies had an amazing run!
To pay respect where respect is due, let’s just get the obvious out of the way: cryptocurrencies had very impressive returns. I downloaded the crypto return series I could quickly retrieve from the web: the S&P Cryptocurrency LargeCap Index and three individual coins: Bitcoin, Ethereum, and Litecoin, with the data coming from the St. Louis Fed “FRED” database. In the chart below is the cumulative (nominal) return since 2017. (Note that the S&P Crypto index only starts on 2/28/2017). The cryptocurrencies had spectacular returns. Ethereum with 360x, Bitcoin at about 41x, Litecoin at 24x, and the Large-Cap crypto index with 20x (though with a slightly later starting date of 2/28/2017). I had to display the y-axis in logs because otherwise, you wouldn’t even notice the S&P 500 stock index and the 10-year benchmark bond index (Source: www.spglobal.com). The S&P 500 merely doubled in value (and that’s not even CPI-adjusted), and the 10-year Treasury Benchmark bond index is essentially flat over the 5+ years.
A caveat: those higher returns came at the price of a lot more risk!
Even the spectacular and sky-high past returns look a little bit more down-to-earth when factoring in the risk you have been taking with the crypto investment. In the table below I display the annualized return stats – annualized return, annualized volatility (= monthly vol times sqrt(12)), and the Sharpe Ratio – and again compare and contrast them with the S&P 500 and 10-year Treasury benchmark bond returns. I display the results for three different time periods: 1) the entire interval since 2017 and the pre- and post-pandemic period.
Average annualized returns for crypto assets were in the high double-digits and even triple digits while the S&P 500 (TR) returned “only” about 16.5% since 2017. Nominal bonds only 2% until the end of March 2022. But the cost of double and triple-digit returns is the double and triple-digit risk inherent in the crypto market. And due to the extraordinary volatility numbers the Sharpe Ratios, i.e., the risk-adjusted excess returns over a risk-free benchmark, look not too different from the S&P 500 over the 2017-2022 time span. Ethereum and Bitcoin did a little bit better than the S&P 500, especially during the second half of the sample. Litecoin did worse than the S&P 500.
The crypto return engine started sputtering in 2021
Also evident in the cumulative return chart: all assets have been under pressure since late 2021. I am not saying that a temporary breather like we’ve seen over the last 6 months is a dealbreaker and automatically makes crypto a bad investment. Heck, stocks and bonds are going through the same phases all the time, including right now. What’s a bit more worrisome is that while stocks are holding up OK right now (a drawdown of about 10% as of April 25, 2022), the two major crypto coins are down 40% (and so is the large-cap index) and the smaller Litecoin is down 70%. Ouch!
What’s more concerning than the 40% crypto drawdown is that crypto has also become a lot more correlated with equities, which brings me to the next point:
Crypto has become “Equities on Steroids”
Occasional drawdowns shouldn’t be too much of a concern if that asset is uncorrelated with the rest of the portfolio. And for the longest time, that was certainly the case with the cryptos. But if I calculate the correlations over the three samples again (full sample, 2017-2019, 2020-2022) we notice that the correlation between the crypto assets and equities has noticeably risen, see the table below:
I can also confirm that the higher correlation is not merely due to the 2020 pandemic volatility. In the chart below, I plot the 12-months rolling correlation with equities. The strong correlation still holds up after rolling out the pandemic bear market. Anything with a market cap approaching $1t will start having some correlations with the business cycle and other risky assets!
Even worse than the correlation, the “betas”, i.e., factor model regression slopes of the crypto returns vs. the equity returns have been substantial. In the table below I display the factor model regression results, i.e., regress crypto (excess) returns on (excess) returns of stocks and bonds. In the second part of the sample, the index and all three coins have substantial and significant beta exposure to the stock market. For example, the 3.25 equity beta and 1.81 bond beta for Ethereum signal that this coin behaves close to a 3.25x leveraged equity plus 1.81x bond portfolio. But I also concede that in addition to the factor exposure, there are significant intercepts (alphas) in all of the coins. Even significant in the case of Ethereum. Though I doubt that Ethereum will keep paying an 11.7% (monthly!) excess return over a 3.25x equity portfolio forever!
But does this all mean that cryptocurrencies are a bad investment? Not necessarily. It depends on the outlook for crypto returns going forward. This brings me to the next section…
What are appropriate expected returns for cryptocurrencies?
If we believe that Bitcoin and Ethereum will again return 127% or 300% annualized, respectively, over the next few years then we’re done. No analysis is needed, then. But I think we can all agree that this kind of run cannot continue forever. With a current crypto market capitalization of $1t, it would only take a few more years before crypto assets are worth more than everything else on earth. Herb Stein’s law comes to mind.
So, what would be an appropriate expected return for the crypto assets? Well, just purely from the beta exposures to the stock and bond market index data, we could set the expected return of the crypto-assets the way you’d do in any other factor model application. For example, if Ethereum has an equity beta of 3.25 and a bond beta of 1.81, then the expected excess return is the beta-weighted sum of stock&bond returns, though adjusted by the cash rate for the portion of the total beta exceeding 1.0. That’s because if you invest in stocks or bonds on margin you have to account for the margin rate and subtract that again.
Then, if we calibrate the inputs as:
- Stock expected returns 8% (3% inflation plus 5% real return),
- 10-year Treasury bond expected returns 3%,
- and the cash/money-market rate as 2.5% over the next 10 years,
… then we get the following expected returns:
- Bitcoin: 14.5%
- Ethereum: 21.3%
- Litecoin: 19.2%
- S&P Large Cap Crypto Index: 16.1%
So, even without including any additional alpha, these are very impressive expected returns. If we assume a current market cap of $1t in the large-cap crypto market and a 16.1% overall index growth, then we should expect almost $4.5t in market cap in 10 years. Pretty impressive. With those expected returns, what kind of crypto portfolio weights can we justify? That brings us to the next section…
Adding Crypto in an “Efficient Frontier Analysis”
With the variance-covariance matrix constructed from the post-2020 return data and the forward-looking return data, we have all the tools necessary to construct efficient frontiers with and without crypto assets. I start with the traditional assets only (stocks&bonds), then add the SP Global Large-Cap Crypto Index, and then all four crypto assets (index + 3 individual coins). The efficient frontiers are in the chart below.
A technical note for the finance purists: The efficient frontiers should stop at the min-vol portfolio and not continue below in the south-easterly direction. Lowering return and increasing risk is clearly not efficient. At this point, I leave that portion, but normally I would plot that “inefficient” as a dotted line. The charts were made with Python instead of my usual Matlab/Octave, so please bear with me
I can also look at the portfolio allocation for different expected return targets, see below. Crypto assets don’t enter the portfolio until about 5.3% expected return, and even then we only add them very slowly. Only for 8.75% and higher target returns would you start adding the crypto assets aggressively. That makes perfect sense because you can get only a max return of 8% from the traditional asset mix. But if you’re targeting any kind of expected return similar to a 100% equity portfolio, there’s very little crypto in the portfolio. Maybe a 4-5% share.
Of course, the crypto assets are indeed useful if you’re interested in going beyond the “normal” range. But how much of the “interesting range”, say, 20% and less volatility, is really improved by the crypto assets? Let’s zoom in and look at the chart below. Not much of an improvement. The efficient frontiers with crypto are essentially right on top of the green parabola that uses only stocks and bonds. Crypto doesn’t help you if you demand a portfolio with a volatility less than a 100% equity portfolio.
What if we increase the crypto expected returns?
The problem with any kind of portfolio optimization method, including efficient frontiers, is that the results are very sensitive to the expected returns we use. I don’t want to be accused of low-balling the expected return assumptions. Especially in light of the large alpha estimates in the factor model regression above. So, how about I slap on an additional 10% expected return to all the crypto-assets. Maybe there’s an additional risk factor that I’m not capturing and the 10% extra returns account for that. But a 25-30% annualized return seems unrealistic over the medium-term because I don’t quite see how the crypto market cap can increase by a factor of almost 14 over the next 10 years (1.3^10=13.79). So, under those unrealistic assumptions, crypto assets will indeed make a noticeable difference in the efficient frontiers, even in the sub-20% risk region. But I don’t believe expected returns are that high.
I can also plot the weights of the various assets as a function of the portfolio expected return again. Targeting a 10% annualized expected return, you’ll indeed include about 20 % crypto-assets.
But again, I don’t find the expected return assumptions very realistic.
Risk Parity and Min-Vol portfolios
If you’re troubled by the way I calibrated the expected returns, there are also a few portfolio construction ideas that work entirely off of the variance-covariance matrix (VCV), thereby sidelining the expected return inputs. Min-Vol portfolio and Risk Parity come to mind. Can crypto assets play a role here? Well, let’s take a look:
Risk Parity targets portfolio weights to equalize what different asset classes contribute to the overall portfolio variance. By the way, there is nothing special about parity, i.e., setting the contributions all exactly equal. We can solve for any target risk contribution share. For example, if we want to use Stocks, Bonds, Bitcoin, Ethereum, Litecoin, then literal Risk Parity would be nonsensical because the three crypto-assets would then contribute three-fifths (60%) of the risk. Rather, I like to keep the contributions from the major asset classes equal. So I target the following splits:
- Traditional assets only, i.e, 50% risk contribution from stocks and bonds
- Add the Crypto Index, and each asset class gets a 1/3 contribution
- Add the three individual crypto coins: then each gets 1/9
- Add only the two major coins, and they both contribute 1/6 of the risk
I display the Risk Parity portfolios in the table below. We start with the well-known result that risk parity in a stock/bond world would necessitate a huge shift to bonds because stocks are so volatile. Adding the crypto assets we’d need about 4.42% to 4.67% of crypto-assets to capture 33.3% of the portfolio risk.
Min-Vol portfolio. In the Stock+Bond world, you’ll end up with an 18% Stock and 82% bond portfolio. Adding crypto assets long-only that’s still the same, implying that you wouldn’t touch crypto-assets even if you had access to them. In a long-short world, you’d short the three individual coins but go long the broader crypto index. The net weight of the crypto-assets is just about -1%. In other words, you’d use the strong equity-crypto correlation to hedge out some of your equity risk through a net short-crypto position. Crypto isn’t useful at all in a min-vol world, but who’s really surprised about that one if these coins have a 100% annual volatility, right?!
Seeking high returns and high risk? There might be a better way!
But what about the folks who are comfortable with higher risk if that means higher expected returns. If you’re still accumulating assets and you can use the Dollar-Cost-Averaging effect, then you can clearly ramp up the risk target and use crypto assets, while walking up the efficient frontier, right? True, but there might be a better way. As I wrote in a post almost 6 years ago, we could take the efficient frontier diagram, identify the Maximum Sharpe Ratio portfolio, and use leverage to expand the efficient frontier to the left. For example, in the base case scenario with the calibrated expected returns, the tangency point is at about 53% bonds and 47% stocks. No crypto assets are used in the maximum Sharpe Ratio portfolio. If we simply extrapolate along this tangency line we can reach points far superior (i.e., lower risk and/or higher return) than any of the efficient frontier using crypto assets. For example, with a 4x leverage (about 212% bonds, 188% stocks, and -300% cash), we’d reach about 13.1% expected return and 32.6% expected risk. Close to the Bitcoin expected return but less than half the risk of Bitcoin. Who needs Bitcoin, then?
In the crypto debate, you often hear the crypto critics warning of the impending doom, while the crypto fans extrapolate the spectacular past returns into the future. Both sides of the argument are “conditionally correct”, i.e. if crypto is all just a bubble, then stay away. And if crypto keeps going up at 100% a year then that’s very attractive. But in today’s post, I wanted to show that cryptocurrencies aren’t that attractive even in the “intermediate case” where you assume that crypto assets have expected returns of “only” around 2x to 3x that of equities. That’s because crypto volatility is simply too high, coupled with a noticeable equity correlation.
The rationale here reminds me of my old post from 2017 (“This fund returned over 100% year-to-date. I’m still not buying it!“) with a similar flavor: an ETF had strong returns, but it still was not worth it. Because the volatility and downside risk, as well as the equity correlation, were complete dealbreakers. (and sure enough, just a few months after publishing the post, the ETF indeed blew up in February 2018, though I’m not necessarily predicting the same for the cryptocurrencies).
So, if you want to mix a few % of Bitcoin into your portfolio, be my guest. But I would not invest in crypto on a large scale. Especially not in retirement when volatility and drawdowns can pose a real headache.
So, take it for what this is: I’m just a personal finance blogger with an opinion and some technical training and tools to test my hypotheses. People may accuse me of being a crypto curmudgeon because I missed out on the great wave up. Maybe that’s true. People have a tendency to justify their past actions and past mistakes. But I hope that even the crypto fans have gotten something out of my work. Use my ramblings as a devil’s advocate argument. Can you convince me and yourself that my analysis is wrong?
Thanks for stopping by today. Looking forward to your comments and suggestions!
Title picture source: History Channel