Mean Reversion During the COVID Crash
The 2020 COVID-19 pandemic caused extreme market volatility. Mean reversion strategies experienced immense pressure. Understanding their performance then provides valuable insights. The S&P 500 (SPX) fell 33.9% from February 19, 2020, to March 23, 2020. This marked the fastest bear market entry in history. Liquidity vanished. Correlations surged. Traditional mean reversion models assume stable market conditions. The COVID crash invalidated these assumptions.
A simple mean reversion strategy buys oversold assets and sells overbought assets. Consider a pair trading strategy. This strategy identifies two historically correlated assets. It buys the underperforming asset and sells the outperforming one when their spread diverges. During the COVID crash, correlations between assets often failed. Many assets moved together, driven by fear. This "risk-off" environment penalized strategies relying on relative value. For instance, airline stocks like Delta Air Lines (DAL) and United Airlines (UAL) typically show strong correlation. Their spread often reverts. During the crash, both dropped simultaneously. Shorting one and longing the other became less effective for relative value capture.
Strategy Performance: A Case Study
Examine a common mean reversion indicator: the Relative Strength Index (RSI). A strategy buys when a stock's 14-day RSI drops below 30. It sells when RSI rises above 70. Consider Apple (AAPL) during the crash. On February 19, 2020, AAPL closed at $80.91 (split-adjusted). By March 23, 2020, it closed at $56.09. Its 14-day RSI dipped below 30 multiple times.
- February 27, 2020: AAPL's 14-day RSI reached 27.6. A buy signal. Price: $66.97.
- March 12, 2020: AAPL's 14-day RSI reached 24.1. Another buy signal. Price: $60.91.
- March 16, 2020: AAPL's 14-day RSI reached 21.3. Another buy signal. Price: $59.60.
A strategy blindly buying these dips would have incurred substantial drawdowns. The initial buy on February 27 saw a further 16.4% decline by March 23. This shows the danger of catching a falling knife. Mean reversion strategies need careful risk management during extreme events. Position sizing and stop-loss orders become paramount.
Conversely, some mean reversion strategies performed well during the subsequent rebound. The market bottomed on March 23, 2020. From March 23 to April 8, 2020, the SPX rallied 24.8%. This swift rebound offered chances for strategies that bought the severe dip. A strategy buying the SPY ETF on March 23 at $222.95 and selling on April 8 at $274.00 would have captured a 22.9% gain. However, few strategies possess the accuracy to time such bottoms without significant risk.
Liquidity and Market Structure Impact
The COVID crash revealed weaknesses in market structure. Liquidity, the ease of buying or selling an asset without affecting its price, disappeared. Bid-ask spreads expanded dramatically. For highly liquid assets like SPY, the average bid-ask spread in February 2020 was around 0.01%. In March 2020, it spiked to 0.05% or higher on volatile days. For less liquid assets, the effect was even greater.
Wider spreads increase transaction costs. This directly reduces mean reversion strategy profits. Strategies relying on frequent, small trades become unprofitable. High-frequency mean reversion, for instance, suffered. Their models, calibrated for tighter spreads, generated false signals or executed at unfavorable prices.
Circuit breakers activated often. On March 9, 12, 16, and 18, 2020, market-wide circuit breakers halted trading. These halts prevent further declines but also freeze liquidity. Mean reversion strategies could not execute trades during these periods. This prevented them from capitalizing on perceived oversold conditions immediately. It also prevented them from exiting positions.
The "dash for cash" phenomenon further impacted markets. Investors sold assets indiscriminately to raise cash. This drove correlations towards 1.0. Even historically uncorrelated assets moved in lockstep. This environment renders traditional mean reversion, which thrives on relative mispricing, ineffective. The fundamental relationships underpinning many mean reversion models broke down.
Adapting Mean Reversion for Extreme Regimes
Successful mean reversion traders used adaptive strategies. They incorporated regime filters. A regime filter identifies current market conditions. During high volatility, a filter might reduce position sizes. It might widen entry/exit thresholds. It might even stop trading entirely.
Volatility-adjusted position sizing holds great importance. Instead of fixed position sizes, strategies dynamically adjust based on market volatility. If the VIX index, a measure of implied S&P 500 volatility, spikes above 40, a strategy might halve its exposure. During the COVID crash, VIX reached 82.69 on March 16, 2020. A static position sizing approach would have led to outsized losses.
Dynamic stop-loss mechanisms also proved essential. Fixed stop-losses, suitable for normal markets, often triggered prematurely during the crash. Volatility-based stops, such as a multiple of the Average True Range (ATR), adapt to market conditions. If ATR expands, the stop-loss widens, giving positions more room to breathe.
Furthermore, some mean reversion strategies benefit from incorporating macro factors. During the COVID crash, news regarding vaccine development, government stimulus packages, and infection rates drove market sentiment. Strategies that could integrate these factors, even qualitatively, performed better. For example, a strategy might avoid buying "oversold" travel stocks if news indicated prolonged lockdowns.
Consider a mean reversion strategy trading the spread between two highly correlated tech stocks, like Microsoft (MSFT) and Apple (AAPL). During normal times, their spread might revert to a mean. During the COVID crash, both stocks initially fell together. A regime filter detecting high VIX or extreme daily SPX moves would have reduced or eliminated trading this pair. This prevents significant losses from a breakdown in correlation.
Backtest mean reversion strategies specifically against extreme market events like the 2020 COVID crash. Do not rely solely on average historical performance. Implement robust regime filters, dynamic position sizing, and adaptive stop-loss orders. These measures reduce risks when market assumptions fail.