Momentum Investing

Can advanced regression techniques identify monthly cross-industry lead-lag return relationships that usefully indicate an industry rotation strategy? In their January 2018 paper entitled “Dynamic Return Dependencies Across Industries: A Machine Learning Approach”, David Rapach, Jack Strauss, Jun Tu and Guofu Zhou examine dynamic relationships between past and future returns (lead-lag) across 30 U.S. industries. To guard against overfitting the data, they employ a machine learning regression approach that combines a least absolute shrinkage and selection operator (LASSO) and ordinary least squares (OLS). Their approach allows each industry’s return to respond to lagged returns of all 30 industries. They assess economic value of findings via a long-short industry rotation hedge portfolio that is each month long (short) the fifth, or quintile, of industries with the highest (lowest) predicted returns for the next month based on inception-to-date monthly calculations. They consider three benchmark hedge portfolios based on: (1) historical past average returns of the industries; (2) an OLS-only approach; and, (3) a cross-sectional, or relative, momentum approach that is each month long (short) the quintile of industries with the highest (lowest) returns over the past 12 months. Using monthly returns  for 30 value-weighted U.S. industry groups during 1960 through 2016, they find that:

Keep Reading

The home page, “Simple Asset Class ETF Momentum Strategy” (SACEMS) and “Simple Asset Class ETF Value Strategy” (SACEVS) now show preliminary positions for February 2018. For SACEMS, past returns for the first and second positions and for the third and fourth positions are close, such that rankings could change by the close. For SACEVS, allocations are unlikely to change by the close.

An anomaly in the source data surfaced this month. Returns for December 2017 for dividend-paying ETFs changed between the end of December 2017 and the end of January 2018. It appears that data available as of the December market close did not account for dividend ex-dates during December. This anomaly has two implications:

1. December 2017 returns previously reported for SACEMS and SACEVS (and alternatives using dividend paying ETFs) were too low. We are correcting these returns.
2. More seriously, incorporation of December 2017 dividends causes a change in the SACEMS top three winners for December 2017, which we determine based on total returns. Since the historical SACEMS performance we present is based on fully updated backtests, the data anomaly introduces a disconnect between backtest and live portfolio performances. In this case, the backtest performs better than a live portfolio. If this issue recurs, we will consider other data management approaches.

Recall the prior data instability reported in “Simple Asset Class ETF Momentum Strategy Data Changes”. Over the long run, data instability issues may cancel with respect to live portfolio performance.

How, in a nutshell, do momentum investing strategies work? In his December 2017 paper entitled “Keep Up the Momentum”, Thierry Roncalli summarizes the nature of the momentum premium in a less mathematical way than in the previously available “Understanding the Momentum Risk Premium: An In-Depth Journey Through Trend-Following Strategies”. He distinguishes between:

• Time-series or trend-following or intrinsic or absolute momentum (long assets with a positive past trend and short assets with a negative past trend).
• Cross-sectional or relative or winners-minus-losers momentum (long assets that have outperformed and short assets that have underperformed relative to each other).

Based on mathematical derivations and prior research, he concludes that: Keep Reading

Subscribers have suggested an alternative approach for the “Simple Asset Class ETF Momentum Strategy” (SACEMS) designed to suppress trading by holding past winners until they fall further in the rankings than in the baseline specification. SACEMS each month picks winners from the following set of exchange-traded funds (ETF) based on total returns over a specified lookback interval:

PowerShares DB Commodity Index Tracking (DBC)
iShares MSCI Emerging Markets Index (EEM)
iShares MSCI EAFE Index (EFA)
SPDR Gold Shares (GLD)
iShares Russell 2000 Index (IWM)
SPDR S&P 500 (SPY)
iShares Barclays 20+ Year Treasury Bond (TLT)
Vanguard REIT ETF (VNQ)
3-month Treasury bills (Cash)

There are three versions of SACEMS: (1) top one of the nine ETFs (Top 1); (2) equally weighted top two (EW Top 2); and, (3) equally weighted top three (EW Top 3). To test the suggestion, we specify three “sticky” versions of SACEMS as follows:

1. Top 1 Sticky – retains the past winner until it drops out of the top 2.
2. EW Top 2 Sticky – retains past winners until they drop out of the top 3.
3. EW Top 3 Sticky – retains past winners until they drop out of the top 4.

We compare sticky and baseline strategies using the tabular performance statistics used for the baseline. Using monthly total (dividend-adjusted) returns for the specified assets during February 2006 (limited by DBC) through December 2017, we find that:

Keep Reading

What is the best way to think about reliabilities and risks of various anomaly premiums commonly that investors believe to be available for exploitation? In their December 2017 paper entitled “A Framework for Risk Premia Investing”, Kari Vatanen and Antti Suhonen present a framework for categorizing widely accepted anomaly premiums to facilitate construction of balanced investment strategies. They first categorize each premium as fundamental, behavioral or structural based on its robustness as indicated by clarity, economic rationale and capacity. They then designate each premium in each category as either defensive or offensive depending on whether it is feasible as long-only or requires short-selling and leverage, and on its return skewness and tail risk. Based on expected robustness and riskiness of selected premiums as described in the body of research, they conclude that: Keep Reading

What is the best way to implement futures momentum and manage its risk? In their November 2017 paper entitled “Risk Adjusted Momentum Strategies: A Comparison between Constant and Dynamic Volatility Scaling Approaches”, Minyou Fan, Youwei Li and Jiadong Liu compare performances of five futures momentum strategies and two benchmarks:

1. Cross-sectional, or relative, momentum (XSMOM) – each month long (short) the equally weighted tenth of futures contract series with the highest (lowest) returns over the past six months.
2. XSMOM with constant volatility scaling (CVS) – each month scales the XSMOM portfolio by the ratio of a 12% target volatility to annualized realized standard deviation of daily XSMOM portfolio returns over the past six months.
3. XSMOM with dynamic volatility scaling (DVS) – each month scales the XSMOM portfolio by the the ratio of next-month expected market return (a function of realized portfolio volatility and whether MSCI return over the last 24 months is positive or negative) to realized variance of XSMOM portfolio daily returns over the past six months.
4. Time-series, or intrinsic, momentum (TSMOM) – each month long (short) the equally weighted futures contract series with positive (negative) returns over the past six months.
5. TSMOM with time-varying volatility scaling (TSMOM Scaled) – each month scales the TSMOM portfolio by the ratio of 22.6% (the volatility of an equally weighted portfolio of all future series) to annualized exponentially weighted variance of TSMOM returns over the past six months.
6. Equally weighted, monthly rebalanced portfolio of all futures contract series (Buy-and-Hold).
7. Buy-and-Hold with time-varying volatility scaling (Buy-and-Hold Scaled) – each month scales the Buy-and-Hold portfolio as for TSMOM Scaled.

They test these strategies on a multi-class universe of 55 global liquid futures contract series, starting when at least 45 series are available in November 1991. They focus on average annualized gross return, annualized volatility, annualized gross Sharpe ratio, cumulative return and maximum (peak-to-trough) drawdown (MaxDD) as comparison metrics. Using monthly prices for the 55 futures contract series (24 commodities, 13 government bonds, 9 currencies and 9 equity indexes) during June 1986 through May 2017, they find that:

Keep Reading

How efficiently can a sophisticated fund manager implement long-only stock momentum portfolios? In their December 2017 paper entitled “Implementing Momentum: What Have We Learned?”, Adrienne Ross, Tobias Moskowitz, Ronen Israel and Laura Serban use seven years of live data for long-only U.S. and international momentum funds to measure the import of implementation frictions. They segment these frictions into turnover/trading costs, tax impacts and mitigating portfolio construction choices. The underlying momentum strategies converge on the top third of stocks based on a combination of market capitalization and momentum signal strength (using multiple measures of momentum), reformed monthly. Portfolio construction employs a transaction cost model to minimize costs by: substituting stocks with similar momentum that are cheaper to trade, trading patiently and employing algorithmic trading rules designed to suppress price impacts of trades. Using detailed trade and performance data for the specified momentum funds during July 9, 2009 through December 31, 2016, they find that: Keep Reading

Do investors underestimate the adverse import of large left tails for future stock returns? In their November 2017 paper entitled “Left-Tail Momentum: Limited Attention of Individual Investors and Expected Equity Returns”, Yigit Atilgan, Turan Bali, Ozgur Demirtas and Doruk Gunaydin investigate the relationship between left-tail risk and next-month returns for U.S. and international stocks. They measure left-tail risk at the end of each month via either of:

• Value-at-risk (VaR) – daily return of a stock at the first (VAR1) or fifth (VAR5) percentile of its returns over the past one year (250 trading days).
• Expected shortfall – average daily return of a stock for the bottom 1% (ES1) or bottom 5% (ES5) of its returns over the past year (250 trading days).

They then sort stocks into tenths (deciles) based on left-tail risk and examine variation in next-month average gross returns across deciles. Using daily prices and monthly firm characteristics and risk factors for U.S. stocks with month-end prices at least $5 during January 1962 through December 2014, they find that: Keep Reading

Do firms that acquire patents in similar technologies persistently perform similarly? In the October 2017 draft of their paper entitled “Technology and Return Predictability”, Jiaping Qiu, Jin Wang and Yi Zhou examine monthly performance persistence of stocks grouped by similarity in recent firm patent activity. Specifically, they:

1. Record the patent activity of each firm by patent class over the most recent three calendar years.
2. Quantify similarity of this patent activity for each pair of firms.
3. Segregate firms into innovation groups based on patent activity similarity (top fifth of quantified similarities).
4. For each month during the next calendar year:
   • Rank stocks into fifths (quintiles) based on average prior-month, similarity-weighted return of their respective groups.
   • Form a hedge portfolio that is long (short) the equal-weighted or value-weighted stocks in the highest (lowest) return quintile.

They focus on gross average monthly return and stock return factor model alphas of the hedge portfolio as evidence of firm innovation group performance persistence. Using firm patent information by technology class during 1968 through 2010, and monthly stock data, quarterly institutional holdings and analyst coverage for a broad sample of U.S. stocks priced greater than $1 during 1968 through 2011, they find that:

Keep Reading

Do strongly accelerating firm earnings identify future outperforming stocks? In the October 2017 revision of their paper entitled “Earnings Acceleration and Stock Returns”, Shuoyuan He and Ganapathi Narayanamoorthy investigate the power of earnings acceleration (quarter-over-quarter change in earnings growth, which is year-over-year change in quarterly earnings) to predict abnormal stock returns. They test a hedged trading strategy that long (short) the equal-weighted tenth, or decile, of stocks with the highest (lowest) earnings acceleration for two holding intervals: (1) starting two days after earnings announcement and ending on day 30; and, (2) starting two days after earnings announcement and ending one day after the next quarterly earnings announcement. They allocate new earnings accelerations to deciles based on the prior-quarter distribution of values of earnings acceleration. They define abnormal return as that in excess of the capitalization-weighted market return. Using quarterly firm characteristics and earnings data and daily returns for a broad sample of U.S. stocks, excluding financial and utility stocks, during January 1972 through December 2015, they find that: Keep Reading