Are strategies that exploit return autocorrelation good places to look for complementary (diversifying) return streams? In the March 2017 version of their paper entitled “Momentum and Covered Calls almost Everywhere”, Stephen Choi, Gil-Lyeol Jeong and Hogun Park examine trend following and covered call strategies at the asset class level both separately and in combination. Their asset class universe consists of three equity indexes, three bond indexes, three commodity indexes and one real estate investment trust (REIT) index. Their trend following (or time series momentum) strategy, which exploits positive autocorrelation of monthly index returns, is long (short) an index when its end-of-month level is above (below) its 12-month simple moving average. Their covered call strategy, which exploits negative autocorrelation (reversion) of index returns, is continuous, such as specified for the CBOE S&P 500 BuyWrite Index. They compare trend following and covered call strategies, separately and in combination, with buy-and-hold for single-class indexes and for multi-class portfolios of indexes. They consider three ways to construct multi-class portfolios (see “Tests of Strategic Allocations Based on Risk Metrics”): (1) maximum diversification (MDR), which maximizes the ratio of the sum of volatilities for individual assets divided by overall portfolio volatility; (2) equal risk contribution (ERC), a form of risk parity with adjustments for correlation; and, (3) equal weight (EW). They rebalance these portfolios quarterly, with volatility/correlation inputs for MDR and ERC based on a 3-year rolling window of historical data. They focus portfolio testing for only 10 years (2007-2016) based on availability of data for covered call indexes. Using the specified data as available from the end of 1971 through 2016, they find that:
- Over available full sample periods for 10 asset classes, the trend following strategy:
- Outperforms buy-and-hold based on annualized gross Sharpe ratio for half the asset class indexes.
- On average across the 10 asset classes, slightly outperforms buy-and-hold based on annualized average gross return (5.6% to 4.9%) and annualized gross Sharpe ratio (0.34 to 0.30).
- Over available full sample periods (different from those available for trend following tests) for six asset classes, the covered call strategy:
- Outperforms buy-and-hold (trend following) based on annualized gross Sharpe ratio for five of six (four of six) asset class indexes.
- On average across the six asset classes, outperforms buy-and-hold based on annualized average gross return (3.5% to 2.0%) and annualized gross Sharpe ratio (0.19 to 0.09).
- On average across the six asset classes, outperforms trend following based on both annualized average gross return (3.5% to 2.7%) and annualized gross Sharpe ratio (0.19 to 0.12).
- Trend following strategy returns exhibit mostly negative or near zero return correlations with cover call strategy returns across six asset classes for which both are available, suggesting opportunities for diversification.
- Substituting trend following strategy and/or (if available) covered call strategy returns for the simple index (buy-and-hold) returns in the three multi-class portfolios generally improves risk-adjusted performance. Specifically, annualized gross Sharpe ratios for MDR, ERC and EW allocations are, respectively:
- 0.42, 0.66 and 0.21 when using simple index returns.
- 0.68, 0.74 and 0.39 when using trend following strategy returns.
- 0.75, 0.80 and 0.25 when using covered call strategy returns.
- 0.74, 0.86 and 0.52 when using both trend following and (if available) covered call strategy returns.
In summary, evidence indicates that combining trend following and covered call strategies may offer attractive diversification of return streams based on exploitation of different autocorrelation modes of an underlying index.
Cautions regarding findings include:
- As noted in the paper, the test period of 10 years is short for reliable inference (especially when using a 12-month moving average), and behaviors around the extremely unusual 2007-2009 financial crisis may dominate some findings.
- The use of indexes rather than liquid funds ignores costs of maintaining a liquid fund. These costs vary by asset class, and are likely higher for covered call indexes than simple indexes. See, for example, “Simple Stock Index Option Strategies” for a test of the differences in performance between the S&P 500 BuyWrite Index and exchange-traded funds constructed to track this index.
- Results do not account for costs of shorting indexes as signaled by the trend following strategy or costs of quarterly rebalancing of multi-class portfolios.
- Testing multiple strategies on the same data introduces data snooping bias, such that the best-performing strategy overstates expectations. There may also be snooping bias in selection of parameters (12-month simple moving average for the trend following strategy, 3-year lookback interval for MDR and ERC volatility/correlation estimates.)
