September tends to be a seasonally challenging month for risk asset performance. It coincides with the close of the fiscal year for many real money funds, who may be inclined to take profits on gains (like the ones we’ve seen for stocks in July and August) and/or harvest tax losses. Over the last 5 years (2017-2021), bitcoin has depreciated by an average of 9% during this month.
We are also in uncharted territory with respect to the global macroeconomic environment. In the US, the current regime combines the inflation volatility of 1973-74 with the recessionary elements of 2001 and 2008. Meanwhile, the last time we saw the Fed reduce its balance sheet in 2018, it contributed to a bear market, and the pace of quantitative tightening (QT) is picking up. Add to that, Europe is wrestling with an energy crisis ahead of the winter, and in China, problems in the real estate sector are deepening while the zero covid policy hasn't done the economy any favors. To say that the current environment is uncertain is an understatement.
In that context, alpha generating strategies may be hard to come by, but proponents of trend-following investing suggest time-series momentum is agnostic to economic conditions as well as bull vs bear markets. Indeed, the convex payout profile of this approach relative to a buy-and-hold strategy looks attractive over a 5 year history. But our model suggests that trend following has underperformed in all but two of the last 7 quarters, though that has still been enough to leave the cumulative return deep in positive territory. We evaluate the relative merits of the strategy with respect to the challenges of the current macro environment.
Something you get through
Our base case is that the correlation between crypto performance and traditional market returns will remain high in the short term (through the end of September), as macro factors continue to be the dominant driver across all risk assets. In our view, crypto may fare better than other asset classes as the liquidations in June already cleared out most of the excess risk. Idiosyncratic narratives like Ethereum’s Merge are more difficult to parse near term, but they are necessary to the medium term setup for the asset class.
Positioning looks very clean as well with stablecoin dominance (i.e. the size of the Tether and USD Coin market cap as a percentage of the total crypto market cap) at a very high 12.7% as of August 31. Stablecoin dominance tends to rise when markets are weak and fall when markets rally and cash is deployed. What we’re seeing right now is that a lot of cash is sitting on the sidelines, waiting for investor confidence to return. Once the bearish sentiment is closer to fully priced in the weeks ahead, that could open the door to inflows in 4Q22 in our view.
1. Rise in stablecoin dominance
Soft sentiment
In the meantime, what’s holding market players back are global macroeconomic conditions, which we believe are perhaps the most complex and uncertain they’ve been in a long time. That has put a ceiling on investor sentiment beginning in the second half of August and it may have more room to run.
The volatility in US inflation levels rivals the one seen in 1973-74 and not just because food, housing and energy have been so costly. It’s also complicated by the labor supply vs demand story affecting services prices (and goods to a lesser extent.) That feeds into concerns that “reducing inflation is likely to require a sustained period of below-trend growth” to temper inflation, according to Fed Chair Jerome Powell at the Jackson Hole Symposium in late August. In short, in our view, the Fed doesn’t know exactly how fast nor to what level it will need to hike rates to put inflation on the path to its 2% target. At the same time, liquidity conditions are likely to tighten at a faster pace starting in September, which doesn’t help risk assets.
Elsewhere, the costs of geopolitical tensions in Europe have been rising for months. A halt in Russian deliveries of natural gas could be a precursor to a region-wide recession, as an energy shortage risks becoming a severe wintertime energy crisis for the European Union. On the upside, Europe’s gas stores reached 81.9% full according to Gas Infrastructure Europe (GIE) as of September 3, above the 80% minimum necessary to get through the winter. Moreover, early indications suggest Europe could have a milder-than-average winter. But infrastructure constraints mean that without Russian flows, an economic shock still seems possible if industries shut down and business confidence declines. This could happen in a nonuniform way among countries, posing a potential existential threat to the single market.
Finally, in China, the real estate problems seem to be deepening, and the country may be getting closer to a flash point, particularly with domestic demand still encumbered by a zero-covid policy. The latter is contributing to a slump in real estate purchases and exacerbating the cash flow issues for property developers who lack the funds to complete projects. As real estate makes up ~25% of China’s GDP (including construction, building materials and services), it seems increasingly unlikely that the country will be able to deliver on its forecast of 5.5% YoY economic growth in 2022.
These developments suggest the engines for global growth are sputtering. We are seeing worldwide economic convergence in the wrong direction, which is keeping pressure on markets across almost all asset classes.
Trend following
To mitigate some of the uncertainty, some speculators rely on trend-following strategies that can theoretically earn alpha in both bull and bear markets. Studies suggest these approaches are agnostic to economic conditions, as they attempt to capture behavioral patterns of (initial) overreacting or (delayed) underreacting to new information. In our case, we use time-series momentum strategies (TSMOM) based on an asset’s own past returns. Moskowitz, Ooi and Pedersen (2012) studied the effect on equity indices, bond futures, FX and commodities and found “persistence in returns for one to 12 months that partially reverses over longer horizons.”
2. Payout profile of monthly TSMOM vs market returns of a weighted BTC and ETH basket
We backtested a simplistic version of this strategy for BTC and ETH based on taking long positions when market returns are rising and short positions when market returns are falling. A scatterplot of monthly returns on a market cap weighted basket of BTC and ETH over a 5 year period suggests a convex payout profile when we compare TSMOM returns (based on a rolling 20-day window) and a buy-and-hold (long only) market strategy. This is consistent with previous studies applying TSMOM to other asset classes.
While trend following strategies seem profitable, particularly on a multi-year cumulative basis, we have seen significant variability in its performance in recent quarters. For example, if we look at the return on our weighted BTC and ETH portfolio between early January 2021 until end-August 2022, a long only approach would have netted a loss of 11% compared to a 70% gain using a TSMOM approach. But decomposing the strategy performance suggests a major part of that difference was due to holding a short position in 2Q22, when long positions suffered a massive 94% loss. That can be a challenge as the infrastructure in crypto is not necessarily there to quickly short assets. That said, even if we altered the time series momentum strategy to hold a flat position during downtrends, we would still earn 39% on our portfolio between start-2021 and late-August 2022 instead of 70%.
3. Quarterly performance of TSMOM vs long only market strategy for BTC and ETH
However, those top line figures obscure the sensitivity of crypto trend following strategies to timing and entry/exit levels. Since the start of 2021 (over the last 7 quarters), the time series momentum approach we applied here actually underperformed a buy-and-hold market strategy in all but two quarters: 2Q21 and 2Q22. See chart 3. That actually predates the rise in the correlation between crypto and US stock returns, as the coefficient didn’t cross the 50% threshold until 1Q22. Traditionally, trend-following strategies perform best in low correlation environments, which is the opposite of what we’re currently experiencing. The bear market rally in July and early August has similarly contributed to the current underperformance of TSMOM in 3Q22, though the signal has shifted to being short in September, which we believe corresponds better to the current weak macro conditions.
4. Crypto correlations with US stocks
It’s important to note that we did not consider slippage, bid-ask spreads and commission fees on the profit margins of the strategies above. This can affect returns depending on the turnover frequency of the portfolio. Our model for example had 27 turnovers in 2022 YTD and 40 turnovers in 2021. Also, because of intraday volatility and the fact that cryptocurrency markets trade continuously 24/7, market players need to consider the fixing reference with which to calculate returns. This tends to be 16h UTC in traditional foreign exchange markets, which is consistent with our study on recurring patterns in crypto liquidity, which indicated that daily liquidity tends to be concentrated in the 14-16h UTC time frame, capturing the market opening hours in NY.
Conclusions
Overall, we think the macro dominant trading environment may continue weighing on crypto performance in the very short term. But we expect digital asset prices may begin to stabilize in early 4Q22 as significant pessimism is currently getting priced in, and the incremental newsflow would need to be significantly more negative to make the sentiment any worse in the medium term.
While trend following strategies could be one way to take a less directional approach to crypto investing, the uncertain macro environment has made it particularly challenging recently to isolate the latent effect of past returns on current performance. Over the longer term (12-18 months), these strategies can still pay off, particularly with a diversified basket, but they remain sensitive to transaction costs, slippage and entry-exit levels.
