Equity Premium

How does the fact that many investments can go up more than they can go down interact with their optimal allocations? In their February 2017 paper entitled “What Our Market Return Forecasts Really Mean: Convexity in Equity Returns and its Implications for Investment Sizing”, Victor Haghani and James White examine how return convexity, return forecast and investment sizing tie together. They measure equity return convexity as arithmetic mean of returns over multiple intervals during a sample period minus geometric mean (compound annual growth rate) over the same sample period. They approximate its value generally as one-half the variance (square of standard deviation) of returns measured across the same multiple intervals. In the context of results from an early 2017 survey of 118 experienced finance professionals about expected U.S. stock market performance, they then consider implications of convexity return when applying the Merton rule for optimal allocation sizing (Sharpe ratio divided by the product of investor risk aversion and standard deviation of returns). Based on mathematical formulas and approximations, they conclude that: Keep Reading

How attractive are purified factor portfolios, constructed to focus on one factor by avoiding exposures to other factors? In their January 2017 paper entitled “Pure Factor Portfolios and Multivariate Regression Analysis”, Roger Clarke, Harindra de Silva and Steven Thorley explore a multivariate regression approach to generating pure factor portfolios. They consider five widely studied factors: value (earnings yield); momentum (cumulative return from 12 months ago to one month ago); size (market capitalization); equity market beta; and, profitability (gross profit margin). They also consider bond beta (regression of stock returns on 10-year U.S. Treasury note returns) to examine interest rate risk. They each month reform two types of factor portfolios:

1. Primary – a factor portfolio with weights that deviate simply from market weights based on analysis of just one factor, with differences from market portfolio weights scaled by market capitalization.
2. Pure – a factor portfolio derived from a multiple regression that isolates each factor, ensuring that it has zero exposures to all other factors.

They measure factor portfolio performance based on: average difference in monthly returns between each factor portfolio and the market portfolio; annualized standard deviation of the underlying monthly return differences; 1-factor (market) alpha; and, information ratio (alpha divided by incremental risk to the market portfolio). Using return and factor data for the 1,000 largest U.S. stocks during 1967 through 2016, they find that: Keep Reading

Which stock market factors and stock characteristics contribute significantly to portfolio performance when considered jointly (accounting for interactions) on a net basis (accounting for offsetting trades)? In their February 2017 paper entitled “A Portfolio Perspective on the Multitude of Firm Characteristics”, Victor DeMiguel, Alberto Martin-Utrera, Francisco Nogales and Raman Uppal investigate which of 51 stock factors/characteristics matter on a net basis when considered jointly rather than individually. They focus on three research questions:

1. Which characteristics are jointly significant from a portfolio perspective on an in-sample, gross basis?
2. How does accounting for trading costs (an in-sample, net basis) affect the answer?
3. Can investors exploit net findings out-of-sample?

They first construct single-characteristic hedge portfolios that are long (short) stocks with expected returns above (below) respective cross-sectional averages. They then construct a parametric multi-characteristic portfolio by adding to a benchmark portfolio the linear combination of single-characteristic hedge portfolios that maximizes mean-variance utility. They next determine which single-characteristic portfolio linear coefficients (the parameters) are significantly different from zero and decompose the contribution of each into gross return, risk and trading friction components. They measure the in-sample performance of a portfolio that exploits those characteristics with significant coefficients on a net basis. Finally, they perform an out-of-sample “big-data” strategy test that each month employs a rolling window of the last 100 months to specify the coefficients of the 51 long-short characteristic portfolios and holds the specified multi-characteristic portfolio the next month. They estimate proportional trading frictions for each stock as a function that decreases with each of two variables, market capitalization of the stock and time. Using monthly return and characteristics data for a broad sample of U.S. stocks (an average of about 3,000 per month) during January 1980 through December 2014, they find that: Keep Reading

Does the original 1963-1997 study identifying (Amihud) illiquidity as a stock pricing factor hold in recent data? In their December 2016 paper entitled “Illiquidity and Stock Returns: Cross-Section and Time-Series Effects: A Replication”, Lawrence Harris and Andrea Amato replicate the original research and extend it to 1998-2015 data. As in the prior study, they: (1) each month measure Amihud illiquidity as the annual average ratio of a stock’s daily absolute return to its daily dollar volume; (2) use monthly regressions to relate stock illiquidity to next-month stock returns and other stock/firm characteristics; (3) quantify how next-month and next-year excess equally weighted stock market return varies with average expected (explained by autoregression) and unexpected (not explained by autoregression) stock illiquidity; and (4) compare the explanatory power of Amihud illiquidity to that of other simple illiquidity measures based on the same absolute returns and dollar volumes. Calculations exclude stocks with extreme (top and bottom 1%) illiquidities as unreliable. Using daily return and trading volume data and contemporaneous monthly characteristics for a broad sample of U.S. stocks during 1963 through 2015, they find that: Keep Reading

Do bonds, like equity markets, offer a variance risk premium (VRP)? If so, does the bond VRP predict bond returns? In their January 2017 paper entitled “Variance Risk Premia on Stocks and Bonds”, Philippe Mueller, Petar Sabtchevsky, Andrea Vedolin and Paul Whelan examine and compare the equity VRP (EVRP) via the S&P 500 Index and U.S. Treasuries VRP (TVRP) via 5-year, 10-year and 30-year U.S. Treasuries. They specify VRP generally as the difference between the variance indicated by past values of variance (realized) and that indicated by current option prices (implied). Their VRP calculation involves:

• To forecast daily realized variances at a one-month horizon, they first calculate high-frequency returns from intra-day price data of rolling futures series for each of 5-year, 10-year and 30-year Treasury notes and bonds and for the S&P 500 Index. They then apply a fairly complex regression model that manipulates squared inception-to-date returns (at least one year) and accounts for the effect of return jumps. 
• To calculate daily implied variances for Treasuries at a one-month horizon, they employ end-of-day prices on cross-sections of options on Treasury futures. For the S&P 500 Index, they use the square of VIX.
• To calculate daily EVRP and TVRPs with one-month horizons, subtract respective implied variances from forecasted realized variances.

When relating VRPs to future returns for both Treasuries and the S&P 500 Index, they calculate returns from fully collateralized futures positions. Using the specified futures, index and options data during July 1990 through December 2014, they find that: Keep Reading

Does suppressing unrelated risks from stock factor portfolios improve performance? In their January 2017 paper entitled “Diversify and Purify Factor Premiums in Equity Markets”, Raul Leote de Carvalho, Lu Xiao, François Soupé and Patrick Dugnolle investigate how to improve the capture of four types of stock factor premiums: value (12 measures); quality (16 measures); low-risk (two measures); and, momentum (10 measures). They standardize the different factor measurement scales based on respective medians and standard deviations, and they discard outliers. Their baseline factors portfolios employ constant leverage (CL) by each month taking 100% long (100% short) positions in stocks with factor values associated with the highest (lowest) expected returns. They strip unrelated risks from baseline portfolios by:

• SN – imposing sector neutrality by segregating stocks into 10 sectors before ranking them for assignment to long and short sides of the factor portfolio. 
• CV – replacing constant leverage by each month weighting each stock in the portfolio to target a specified volatility based on its actual volatility over the past three years.
• HB – hedging the market beta of the portfolio each month based on market betas of individual stocks calculated over the past three years by taking positions in the capitalization-weighted market portfolio and cash.
• HS – hedging the size beta of the portfolio each month based on size betas of individual stocks calculated over the past three years by taking positions in the equal-weighted market portfolio and the capitalization-weighted market portfolio.

They examine effects of combining measures within factor types, combining types of factors and excluding short sides of factor portfolios. They also look at U.S., Europe and Japan separately. Their portfolio performance metric is the information ratio, annualized average return divided by annualized standard deviation of returns. Using data for stocks in the MSCI World Index since January 1997, in the S&P 500 Index since January 1990, in the STOXX Europe 600 Index since January 1992 and in the Japan Topix 500 Index since August 1993, all through November 2016, they find that: Keep Reading

Do individual stocks react differently and persistently to aggregate investor mood changes? In their December 2016 paper entitled “Mood Beta and Seasonalities in Stock Returns”, David Hirshleifer, Danling Jiang and Yuting Meng investigate whether some stocks have higher sensitivities to investor mood changes (higher mood betas) than others, thereby inducing calendar effects in the cross-section of returns. They specify mood based on three calendar-based U.S. stock market return anomalies:

1. January (highest average excess return of all months) represents good mood, while October (lowest average excess return of all months) represents bad mood.
2. Friday (highest average excess return of all days) represents good mood, while Monday (lowest average excess return of all days) represents bad mood.
3. The two days before holidays (abnormally high average excess return) represent good mood, while the two days after holidays (abnormally low average excess return) represent bad mood.

They structure their investigation via a factor model of stock returns, with mood as a factor. They measure a stock’s mood beta by regressing its returns during high and low mood intervals versus contemporaneous equal-weighted market returns over a rolling historical window. Each year, they regress a stock’s monthly January and October returns versus monthly equal-weighted market returns for those months over the last 10 years. Each week, they regress a stock’s daily Friday and Monday returns versus contemporaneous equal-weighted market returns for those days over the last ten weeks. Each holiday, they regress a stocks pre-holiday and post-holiday daily returns versus versus equal-weighted market returns for those days over the last year (including the same holiday the previous year. They then use the stock’s mood betas to predict its returns during subsequent times of good and bad mood. Using daily and monthly stock returns for a broad sample of U.S. common stocks during January 1963 through December 2015, they find that: Keep Reading

“Cross-asset Class Intrinsic Momentum” summarizes research finding that past country stock index (government bond index) returns relate positively (positively) to future country stock market index returns and negatively (positively) to future country government bond index returns. Is this finding useful for specifying a simple strategy using exchange-traded fund (ETF) proxies for the U.S. stock market and U.S. government bonds? To investigate we test the following five strategies:

1. Buy and hold.
2. TSMOM Long Only – Each month, hold the asset (cash) if its own 12-month past return is positive (negative).
3. TSMOM Long or Short – Each month, hold (short) the asset if its own 12-month past return is positive (negative).
4. XTSMOM Long Only – Each month hold stocks if 12-month past returns for stocks and government bonds are both positive, and otherwise hold cash. Each month hold bonds if 12-month past returns are negative for stocks and positive for government bonds, and otherwise hold cash. 
5. XTSMOM L-S-N (Long, Short or Neutral) – Each month hold (short) stocks if 12-month past returns for both are positive (negative), and otherwise hold cash. Each month hold (short) bonds if 12-month past returns are negative (positive) for stocks and positive (negative) for bonds, and otherwise hold cash.

We use SPDR S&P 500 (SPY) and iShares 7-10 Year Treasury Bond (IEF) as proxies for the U.S. stock market and U.S. government bonds. We use the 3-month U.S. Treasury bill (T-bill) yield as the return on cash. We apply the five strategies separately to SPY and IEF, and to an equally weighted, monthly rebalanced combination of the two for a total of 15 scenarios. Using monthly total returns for SPY and IEF and monthly T-bill yield during July 2002 (inception of IEF) through December 2016, we find that: Keep Reading

Are stock and bond markets mutually reinforcing with respect to time series (intrinsic or absolute return) momentum? In their December 2016 paper entitled “Cross-Asset Signals and Time Series Momentum”, Aleksi Pitkajarvi, Matti Suominen and Lauri Vaittinen examine a strategy that times each of country stock and government bond (constant 5-year maturity) indexes based on past returns for both. Specifically:

• For stocks, they each month take a long (short) position in a country stock index if past returns for both the country stock and government bond indexes are positive (negative). If past stock and bond index returns have different signs, they take no position.
• For bonds, they each month take a long (short) position in a country government bond index if past return for bonds is positive (negative) and past return for stocks is negative (positive). If past stock and bond index returns have the same sign, they take no position.

They call this strategy cross-asset time series momentum (XTSMOM). For initial strategy tests, they consider past return measurement (lookback) and holding intervals of 1, 3, 6, 9, 12, 24, 36 or 48 months. For holding intervals longer than one month, they average monthly returns for overlapping positions. For most analyses, they focus on lookback interval 12 months and holding interval 1 month. For a given lookback and holding interval combination, they form a diversified XTSMOM portfolio by averaging all positions for all countries. They measure excess returns relative to one-month U.S. Treasury bills. They employ the MSCI World Index and the Barclays Capital Aggregate Bond Index as benchmarks. Using monthly stock and government bond total return indexes for 20 developed countries as available during January 1980 through December 2015, they find that: Keep Reading

How many factors are optimal for modeling future returns of individual stocks? How do these factors relate to conventionally used factors (market, size, value, momentum, investment, profitability…)? In the June 2016 version of their paper entitled “Multifactor Models and the APT: Evidence from a Broad Cross-Section of Stock Returns”, Ilan Cooper, Paulo Maio and Dennis Philip derive mathematically an optimal set of factors for predicting returns of 278 stock portfolios created by sorting U.S. stocks into tenths (deciles) according to 28 market anomalies encompassing aspects of value, momentum, investment, profitability and intangibles. They apply asymptotic principal components analysis to these portfolios to identify the factors. They quantify the premium of each of these factors as the average return spread between extreme deciles of monthly sorts of the 278 source portfolios on the factor. They then examine interactions between this mathematical factor set and several widely used empirical multi-factor models: the Fama-French 3-factor model (market, size, book-to-market); a 4-factor model (adding momentum to the 3-factor model); a second 4-factor model (adding liquidity to the 3-factor-model); a third 4-factor model (market, size, investment, profitability); and, a 5-factor model (adding investment and profitability to the 3-factor model). Using monthly returns for the 278 source stock portfolios during January 1972 through December 2013, they find that: Keep Reading