## 04 Aug Is Low Net Actually Low Risk?

According to capital introduction groups, institutional investor sentiment is trending toward equity long/short (“L/S”) strategies with low net equity exposure (“nee”) (i.e., +/- 0.20). Given the current low market volatility, expectations of increased volatility may be motivating this. I infer that institutional investors view L/S strategies with low nee as providing returns with lower volatility because many investors equate net exposure with equity market exposure (or sensitivity) and market volatility. However, I think this is a mistake.

Some strategies with low nee exhibit a couple of negative traits: (1) negative market sensitivity; and (2) higher volatility than is best given the expected return. To illustrate the point, I present a “buy cheap, sell expensive” L/S strategy that produces less volatile returns by increasing nee and that increases market exposure modestly.

__An Earnings-based Example__

I created an equity long/short strategy based on data from the Kenneth French library. Using the earnings yield monthly returns from July 1951 to Dec. 2014, I created a long portfolio of the top 30% of earnings yield securities and a short portfolio of the securities with negative or zero earnings yield. A performance plot of the strategy (less cash) with nee = 0% and nee = 45% shows that the higher nee series is less volatile and experiences less severe drawdown events.

In fact, at nee = 45%, this strategy exhibits the least volatility of all the scenarios. Consider the following annualized performance comparisons from July 1951 to Dec. 2014:

Return increased 7.93% and volatility decreased 6.94%. It is notable that the market beta increased to 0.26 from -0.34. Assuming that the market moves up more often than it moves down, this changes strikes me as a move in the right direction.

Because of the increased market sensitivity, I think it appropriate to identify the portion of the return attributable to the increase. I multiplied the change in market beta (0.60) by the market return (6.43%), dividing the result by the change in performance (7.93%). The result implies that 48.38% of the performance improvement is attributable to market exposure. As an aside, the same analysis using the S&P 500 shows that only 0.50% of the performance improvement is attributable to increased market sensitivity. The choice of broad market measure is clearly a relevant consideration.

At this point, we have shown that you can increase nee and increase risk-adjusted return with lower volatility and higher return. Also, we show that a subset of the performance increase is due to increased market exposure. Let’s continue to dissect the performance to see if there are any other factors impacting the result.

__Common Factor Exposures__

I thought it prudent to evaluate the sensitivities to common factors. I regressed the returns over the following Fama/French benchmark factors: Market, Value/Growth, Size, and Momentum. The market return is the value-weighted US return in excess of cash.

Equal-Weighted Holdings | |||||

Regression Coefficients | |||||

NEE | Market | Size | Value/Growth | Momentum | Excess Return |

0.00 | 0.22 | -0.24 | 0.35 | 0.15 | 0.52 |

0.45 | 0.32 | -0.08 | 0.38 | 0.14 | 0.24 |

The t-statistics significant at the 1% level are as follows:

NEE | Market | Size | Value/Growth | Momentum | Excess Return |

0.00 | 9.60 | -6.80 | 9.80 | 5.40 | 7.00 |

0.45 | 16.00 | -2.60 | 12.00 | 5.80 | 3.70 |

*Note: Number of observations = 762*

The adjusted R^{2} at nee = 0% and nee = 45% are 0.20 and 0.32, respectively.

The low volatility return fits the model slightly better. But, the increase to the Market coefficient is marginal and possibly related to the reduced excess return.

While the increase in the Market coefficient was not large, the increase in the Size coefficient was notable. The Size factor is the spread between small and large companies by market capitalization (or small minus large). Because the analysis relies on the equal-weighted monthly return series, I repeated the analysis with the market value-weighted monthly returns to observe any changes to the Size coefficient.

The value-weighted version has the lowest volatility at nee = 55%. Like the equal-weighted version, return increased 6.92% and volatility decreased 6.66%. The market beta changed to 0.32 from -0.44.

Regarding the common factor coefficients, Market and Size increased. Similar to the equal-weighted version, return series fits the model better; the adjusted R^{2} for nee = 0% and nee = 55% are 0.40 and 0.46, respectively.

Value-Weighted Holdings | |||||

Regression Coefficients | |||||

NEE | Market | Size | Value/Growth | Momentum | Excess Return |

0.00 | -0.23 | -0.83 | 0.36 | 0.14 | 1.55 |

0.55 | 0.44 | -0.33 | 0.44 | 0.14 | 0.31 |

The t-statistics significant at the 1% level are as follows:

NEE | Market | Size | Value/Growth | Momentum | Excess Return |

0.00 | -6.60 | -16.00 | 6.50 | 3.40 | 14.00 |

0.55 | 22.00 | -11.00 | 14.00 | 5.60 | 4.90 |

*Note: Number of observations = 762*

__Conclusion__

Is lower volatility worth a marginal increase in market exposure? Yes. While nee outside the low net range can increase a L/S portfolio’s sensitivity to the market, there are additional considerations:

1. drawdowns hurt performance

2. changes in common factor exposures are important

3. characteristics of the underlying stocks and their portfolio weightings matter

Over the course of my next few posts, I will dig deeper into the nee question …