Building a Superior SPY Hedge

Last week, we discussed the significant drawbacks of using the S&P 500 SPDR (SPY) as a market exposure hedge. We established that 1) the SPY hedge is wildly common among the institutional investment community (40% of the SPY’s AUM is short interest), and 2) it carries an inherent long bet on Volatility - a negative premia factor that’s down 32% over the past five years.

Volatility Factor Return (Market Neutral-Basket of high vol stocks vs. low vol stocks)

Today we’ll show how we’ve built a better hedge basket using the constituents of the S&P 500 that removes the Volatility overhang, while not deviating far from the rest of the SPY’s exposures. We’ll also provide our weekly market and factor trend update.

‍Capturing the Volatility Premium — Methodology

‍Using only the constituents of the S&P 500 as the universe, we used our optimization engine to create a monthly rebalanced hedge basket with the following objectives and constraints:

• Objective: Maximize exposure to the Volatility factor
• Constraints:
• Maintain all other factor exposures as close as possible to the original SPY exposures
• Portfolio turnover <= 10% on every monthly rebalancing

For the purposes of discussion and comparison below we will refer to our version of SPY as “SPY VOLHEDGE”.

‍SPY VOLHEDGE Composition (Past 5 years)

‍The resulting SPY VOLHEDGE basket has on average 201 securities over the past five years vs. the SPY’s 500, but is far more evenly distributed in weights of the securities in the basket.

Here are the top 10 constituents of each:

Min and Max Concentration: 10/24/14 - 10/24/19

Comparing Factor Exposures: SPY VOLHEDGE vs. S&P 500

As we can see, compared to the SPY, our SPY VOLHEDGE basket maintains similar exposures relative to the SPY save for a few exceptions:

Volatility Our chief goal for this basket was to turn the SPY’s negative Volatility exposure into a positive one, so that our new hedge won’t lose money due to the natural negative performance of the factor. We were able to successfully increase Volatility exposure up to 0.29 (from -0.07) — a +0.36 change. Based on the annualized performance of the Volatility factor (-5.75% per year) this change alone brings a potential +2.07% incremental annualized return to an investor looking to hedge out SPY exposure.

‍Size - Because we want the SPY VOLHEDGE to have higher exposure to Volatility, the basket has a bias to lower market cap securities. Still, since all constituents came from the S&P 500, the smallest cap name in the SPY VOLHEDGE is NKTR at $2.89B.

Liquidity - Our new basket also has higher exposure to liquid names, which isn’t necessarily a bad thing.

Dividend Yield - The SPY VOLHEDGE is overweight dividend yielding names relative to the SPY.

Market Sensitivity - This basket is underweight higher beta names relative to the SPY.

‍Cumulative Performance: SPY VOLHEDGE vs. SPY (10/24/14 - 10/24/19)

Here, we see that the SPY (+71.07%) outperformed our SPY VOLHEDGE (+42.67%) by 28.4% over the past five years. Because these hedges are being used in the short leg of our portfolios, an investor would have seen an incremental +5.68% additional return per year by using the new basket rather than the SPY.

‍Breaking down the -28.4% difference in cumulative performance between SPY & SPY VOLHEDGE into its factor contributors, we observe:

• The Volatility factor we were targeting was the majority contributor (-17.82% or roughly ~62%) of the total incremental performance
• The incremental performance difference carries some noise (shown by the choppy “alpha” line), but is generally negligible compared to the factor contribution. This is our intention as we want to do our best to avoid idiosyncratic situations in the securities we selected (M&A, outsized earnings events).

We hope that this exercise was informative and helpful in demonstrating how factors can be used as a mechanism to optimize a common hedging task within the institutional investor community.


Market and Factor Update

US Market

• The market had another good week, with the S&P 500 hitting its second-highest close ever on Friday (not captured in above chart) on optimism that Washington is close to finalizing parts of a trade deal with China, as well as some better-than-expected 3Q earnings beats.
• Despite lack of consensus among Fed policy makers, the market is pricing in another Fed funds rate cut of 25bps on Oct 30th.
• Intel (INTC) had a strong quarter, buoying the SOX index to a record high close. More major earnings announcements are on the come next week, including Apple (AAPL), Alphabet (GOOGL), Pfizer (PFE), and Merck (MRK).

‍US Model

• Earnings Yield was the biggest winner in the US this week, and has entered Overbought territory at +1.05 standard deviations above the mean.
• Momentum continued to revert from the recent bottom of -2.98 standard deviations below the mean (9/20), and has now shed its Oversold designation.
• Market Sensitivity and Volatility both continued to see solid gains on a normalized basis.
• The rally in Size continued, albeit at a slower pace, and has now earned an Overbought label at +1.02 SD above the mean.
• Profitability fell the most this week, and crossed back into Neutral territory after popping into Overbought space over the past couple of weeks.
• Value continued to fall away from its recent peak of +2 SD above the mean (9/30)
• US Total Risk (using the Russell 3000 as proxy) saw a slight decrease of 12 bps.

‍Worldwide Model

• Volatility had the biggest upwards move globally as it climbed +0.34 standard deviations above the mean.
• Earnings Yield saw continued strength and has now become an Extremely Overbought factor.
• Size also continued to see some strength on a normalized basis, although its rise has slowed a bit vs. the past couple of weeks.
• Momentum also continued to see gains, losing its Oversold designation after hitting a floor of -2.51 SD below the mean on 9/23.
• Exchange Rate Sensitivity was the biggest loser, as it entered negative normalized return space.
• Global Risk (using the ACWI as proxy) drifted down by 10 bps.

Regards,
Omer