Momentum Crashes

As the three and five-factor models failed to explain the returns on momentum strate-gies, there must clearly be a hidden component to the risk which explains the substantial excess returns (Jegadeesh & Titman, 1993 & 2001; Rouwenhorst, 1999). The analogy

“collecting dimes in front of a steamroller” is depictive of the situation at hand when implementing unhedged momentum strategies for long periods of time. The distribution of the returns on momentum strategies is negatively skewed, thus suggesting an in-creased likelihood of extreme negative events. Occasionally equity momentum strate-gies experience severe crashes like those experienced in currency carry-trades. These crashes occur during and after states of panic, mainly following major market meltdowns and times of heightened market volatility (Daniel & Moskowitz, 2016). Additionally, mo-mentum strategies possess inferior performance during recessions. Grobys (2014) finds recessions to reduce the monthly performance of a momentum strategy by more than 2.2%, making the average returns negative.

As discussed earlier, during normal market activity, equity momentum strategies are close to being market neutral (Jegadeesh & Titman, 1993; 2001). However, market expo-sure significantly fluctuates when markets experience heightened uncertainty. During severe bear markets, momentum strategies pose a -0.70 beta towards the market as it is spiralling downwards, thus continuing the impressive gains on the way down (Daniel

& Moskowitz, 2016). The positive performance during market downturns may be ex-plained by the market liquidity. Increasing liquidity affects equity prices positively (Ami-hud, 2002). During market downturns liquidity starts to evaporate which starts a flight to liquidity phenomena, resulting in investors flocking into more liquid stocks. In general, the past winners are far more liquid than the past losers. Because of the latter, the value of the loser portfolio falls more than the value of the winner portfolio, resulting in a positive overall return and thus, a negative beta to the overall market. (Butt & Virk, 2017.)

However, the problem arises as the market reverses and starts recovering. During post-bear market reversals, the momentum exposure to the overall market is -1.51, making

momentum crashes occur when the overall market returns are high (Daniel & Moskowitz, 2016). These crashes can be attributed to the behaviour of the short leg (the loser port-folio). During market reversals, the past 12-month winner portfolio poses significantly lower returns than the overall market. However, after the reversal, the short leg experi-ences significantly higher returns than the overall market, resulting in significant losses from the short positions. (Daniel & Moskowitz, 2016.)

The behaviour of the short leg makes it essentially a written call option during bear mar-kets. The maximum upside on the shorted stocks is limited to 100% which presents the premium received on the option. However, the potential downside on the stocks is sub-stantial as during bear markets, they are more than likely priced for bankruptcy. If a bank-ruptcy does not occur, the rebound on these shorted stocks is violent. During bear mar-kets momentum returns decrease while the overall market variance increases. This clearly supports the notion of the option-like behaviour of the short leg, as rising uncer-tainty increases option prices. (Daniel & Moskowitz, 2016.) Another explanation for the higher returns on the loser portfolio may also be the sharply increasing market liquidity, which inverses the effect that occurred during the downturn (Butt & Virk, 2017).

Overall, this combination causes the zero-cost strategy to crash within a short period of time, wiping out majority of the invested capital in the worst cases. Table 5 presents the ten worst monthly returns for a momentum strategy investing in U.S. equities with re-spective to the underlying market. The rightmost column tracks the past 2-year returns on the underlying market. Apart from 2001, all momentum crashes followed a severe bear market. (Daniel & Moskowitz, 2016.)

Table 3. Momentum crashes (Daniel & Moskowitz, 2016)

The literature clearly suggests that momentum strategies contain a severe crash risk.

Following this, one can argue that the abnormal returns from the strategies should then be corrected for such risk. Ruenzi and Weigert (2018) amplify the Fama-French three and five-factor models by including a fourth (sixth) variable. The variable is the return of an investment strategy that is long on equities with high crash sensitivity and short on eq-uities with low crash sensitivity. The rationale is that investors dislike assets with high sensitivity to crashes, thus increasing the risk premium on them – resulting in higher expected returns (Kelly & Jiang, 2014). The equities included in the loser portfolio pos-sess higher negative skewness, thus increasing risk premia and the probability of a price reversal if the skewness starts normalizing (Ruenzi & Weigert, 2018).

As seen earlier, when first correcting for the Fama-French three-factors, the momentum strategy poses a significant positive alpha (11.9%) over a 50-year sampling period (1963-2012). The alpha remains positive and significant (13.2%) even when correcting for the five factors. However, after controlling for the new source of risk, both alphas reduce significantly (to 1.8% and 2.9%) in addition to becoming indistinguishable from zero. Ad-ditionally, upon the inclusion of the fourth (sixth) variable, the R-squared nearly doubles for both models (from 0.10 to 0.18). (Ruenzi & Weigert, 2018.)

The findings on momentum crashes suggest that majority of the previously unexplained abnormal returns may just be an adequate compensation for the exposure to crash risk.