Mean-ETL Portfolio Construction in US Equity Market

Mean-ETL Optimization in HorseRace Competition

“…Despite the Mean-ETL portfolios with MQ variable perform better than the other two variables across all three universes in this period, we cannot draw the conclusion that the MQ variable outperforms other variables with different portfolio construction techniques or time periods. For example, Mean-ETL portfolio with CTEF variable outperforms the portfolio with MQ variable in the U.S. stock market in Shao [8].…”

“…As summarized in the previous section, we need to have matrix of scenarios of the underlying stocks in the Mean-ETL optimization. In these HorseRace portfolios, we use similar scenarios generation method as previous works by Shao et al [13] and Shao [8]: Autoregressive moving average (ARMA)-generalized autoregressive conditional heteroscedasticity (GARCH) model with the innovations follow multivariate normal tempered stable distribution (MNTS). For the convenience and consistence reason, we use ARMA-GARCH-MNTS to denote this model in this paper.…”

“…Compared to REG10, CTEF, and PM, there have been fewer studies on the efficiency of the MQ variable in stock selection: Guerard et al [18] found the MQ variable produces statistically significant returns in the vast majority of the eight universes: the Russell 3000, MSCI All Country World, MSCI ex-US, MSCI Emerging Markets, MSCI Japan, MSCI China, a broader China universe and a two-analyst I/B/E/S global universe. Shao [8] showed that the Mean-ETL portfolio of the U.S. stocks based on the MQ variable outperforms the portfolio based on the USER and PM variables with highly statistically significant active returns.…”

“…As for the EM stock universe, the annual active return of EM-MQ is 7.67% with an information ratio of 1.09 and a t statistic of 3.77. Other than the three stock universes we have studied in this paper, the recent study of Shao [8] reports that the MQ variable in the domestic (US) stock universe generates an active annual return 9.72% with an information ratio of 1.12 and a t statistic of 4.18 during the period of 2000-2013.…”

“…Mean-ETL portfolio construction on the fundamental variables has been investigated in the past, e.g., Guerard [7] studied the Global Expected Return (GLER) on the global stock market and the United States Expected Return (USER) on the U.S. stock market during the period of 2003-2011; Shao [8] reported the results of using USER, Price Momentum (PM), and Mckinley Capital Quant Score (MQ) variables on the U.S. stock market during the period of 2000-2013. The study is a more comprehensive study on the application of Mean-ETL portfolio construction with three different fundamental variables, and performs on three stock universes for the time period of 2005-2016.…”