Factors are return drivers that carry intrinsic risk. Macro risk factors capture non-diversifiable risks that have exhibited positive expected returns over long run. Such risk factors include economic, real rats, credit, liquidity, inflation and emerging market, etc. (Ang, 2015). An investor holding bonds expose him to the risk of inflation and real rates hike, and is thus compensated for bearing such risks. Another source of return and risk are style risk factors. Such factors (e.g., value, momentum, earning quality, size, volatility, carry, curve, convexity, etc.) capture risk premiums that have been historically positive if held for long term.
I conducted empirical studies on the performance of investment strategies employing various quant equity factors during 1990 to 2016, and statistically significant anomalous profits are found for trading value, momentum, low-volatility and low-liquidity (Roger G. Ibbotson, 2013) portfolios. The main dataset used in the study is the stock return data from the CRSP database and the data available at Ken French’s data library[1]. In this section, we will cover mainly the empirical results of value and momentum portfolio during 1990-2005 and 2012-2016.
Value Portfolios
Investors seek value premium by holding high book-to-market ratio (B/M ratio) firms. The failure of CAPM to explain the returns of value/growth stocks is known as Value Puzzle. Empirically, value stocks outperform growth stocks in general. According to CAPM, the only factor impacting asset returns is beta, and difference in asset returns should be explained entirely by beta spreads. However, the beta spreads are found to be not distinguishable among value and growth portfolios, while their return expectations are significantly different. This is a challenge for CAPM. During 1990 to 2005, value portfolios[2] outperformed growth and market portfolios, in both absolute and risk-adjusted basis, by 2.05% annually. However, the recent performance of growth portfolios during 2012-2016 outperformed that of value portfolios by 1.24%. In addition, the growth portfolio used to be more volatile with larger maximum drawdown compared with value portfolios during 1990-2005, but it turns into a safer strategy with less drawdown during 2012-2016. These characteristic changes all suggest a potential regime change and factor rotation.
Portfolio Construction Methodology: Every June, we sort firms according to their B/M and form 3 portfolios: Value portfolio is constructed using the firms whose B/M ratio is ranked in the highest 30%, Growth portfolio is constructed using companies whose B/M ratio is ranked in the lowest 30%, and Neutral portfolio is constructed using the remaining 40% of firms. The return from July to following June is computed for each of the three style portfolios, and firms are resorted according to the current B/M to form three new style portfolios. This process is conducted from 1990 to 2016, and finally, the cumulative returns of each style portfolio are computed and compared against the returns of the market portfolio. The market portfolio return is obtained by adding back the risk free rate to the market risk premium factor used in the Fama French 3-factor model. All data used in this analysis is obtained from Ken French’s Data Library.
Value Growth Styles Summary, 1990-2005
| STYLE | MEAN | STD | SHARPE | MAXDD | SKEW | CORR |
| Growth | 11.13 | 16.01 | 0.44 | 65.24 | -0.39 | 0.98 |
| Value | 13.18 | 13.75 | 0.67 | 33.41 | -0.63 | 0.82 |
| Market | 11.24 | 14.78 | 0.49 | 55.65 | -0.59 | 1 |
Value Growth Styles Summary, 2012-2016
| STYLE | MEAN | STD | SHARPE | MAXDD | SKEW | CORR |
| Growth | 14.98 | 11.53 | 1.3 | 8.95 | -0.29 | 0.98 |
| Value | 13.74 | 12.74 | 1.08 | 15.14 | -0.32 | 0.92 |
| Market | 14.16 | 11.36 | 1.24 | 9.11 | -0.31 | 1 |
Momentum Portfolios
Momentum effect is more significant than value/ growth, but it can occasionally be trapped at crashes, while value/ growth effects are more stable. Compared with Value and Growth portfolio, momentum (i.e. winner) portfolios[3] have historically provided more significant abnormal returns, which is unexplainable by CAPM. Concretely, from 1990 to 2005, winner portfolio have generated on average 19.49% return annually with a Sharpe ratio of 0.71, skewness of -0.3, excess kurtosis and correlation of 0.81with the market portfolio[4]. In contrast, loser portfolios generated only 2.58% annual return with a Sharpe ratio of -0.05. One can also construct a market neutral winner minus loser (W-L) portfolio, which will generate 16.92% annually with a -0.18 correlation with the broad market. However, such momentum portfolio would have incurred a 60.66% maximum drawdown, which is too risky to be included in institutional quality investment programs. These drawdown episodes of momentum strategy are characterized as momentum crashes, which are usually observed during stressful periods and equity bubbles. During the bullish run of 2012-2016, momentum portfolio continued to be a strong performer. However, given the high valuation of the equity market, increased negative skewness and increased average correlation with the market return, one trading such strategy should be cautious about the systematic risk of a sharp equity drawdown.
Portfolio Construction Methodology: Momentum portfolio is constructed by sorting stocks according to returns over past 12 to 2 months to form deciles. The top 10% firms are purchased into the momentum (winner) portfolio, and the bottom decile firms are purchased into the loser portfolio. This rebalance is performed at the end of each June.
Momentum Deciles Monthly Summary, 1990-2005
| DECILES | MEAN | STD | SHARPE | MAXDD | SKEW | CORR |
| ‘Low’ | 0.21 | 8.47 | -0.05 | 100 | 0.33 | 0.77 |
| ‘Dec_2’ | 0.77 | 6.3 | 0.24 | 76.63 | 0.05 | 0.8 |
| ‘Dec_3’ | 0.76 | 5.11 | 0.29 | 69.26 | 0.04 | 0.8 |
| ‘Dec_4’ | 0.93 | 4.44 | 0.47 | 43.3 | -0.21 | 0.82 |
| ‘Dec_5’ | 0.72 | 3.97 | 0.33 | 26.95 | -0.55 | 0.85 |
| ‘Dec_6’ | 0.83 | 3.94 | 0.43 | 44.49 | -0.45 | 0.84 |
| ‘Dec_7’ | 0.96 | 3.91 | 0.55 | 26.13 | -0.32 | 0.83 |
| ‘Dec_8’ | 1.19 | 3.85 | 0.77 | 27.83 | -0.23 | 0.87 |
| ‘Dec_9’ | 1.05 | 4.17 | 0.6 | 29.72 | -0.25 | 0.86 |
| ‘High’ | 1.62 | 6.29 | 0.71 | 60.66 | -0.18 | 0.81 |
Momentum Deciles Monthly Summary, 2012-2016
| DECILES | MEAN | STD | SHARPE | MAXDD | SKEW | CORR |
| ‘Low’ | 0.5 | 7.18 | 0.24 | 50.4 | -0.23 | 0.72 |
| ‘Dec_2’ | 0.59 | 4.9 | 0.41 | 28.73 | -0.29 | 0.83 |
| ‘Dec_3’ | 1.04 | 4.14 | 0.87 | 17.89 | -0.29 | 0.88 |
| ‘Dec_4’ | 1.16 | 3.37 | 1.19 | 10.46 | -0.43 | 0.94 |
| ‘Dec_5’ | 1.45 | 3.53 | 1.42 | 8.95 | -0.19 | 0.96 |
| ‘Dec_6’ | 1.17 | 3.23 | 1.26 | 8.02 | -0.33 | 0.96 |
| ‘Dec_7’ | 1.28 | 3.16 | 1.4 | 8.87 | -0.43 | 0.96 |
| ‘Dec_8’ | 1.31 | 2.89 | 1.57 | 9.68 | -0.36 | 0.93 |
| ‘Dec_9’ | 1.36 | 3.29 | 1.43 | 8.75 | -0.11 | 0.89 |
| ‘High’ | 1.56 | 3.9 | 1.38 | 10.05 | -0.27 | 0.82 |
[1] http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
[2] The value/growth portfolio construction methodology can be found in Appendix B.
[3] The momentum portfolio construction methodology can be found in Appendix B.
[4] Market portfolio return is obtained by adding risk free rate back to the excess return used in the Fama French 3-factor model.
