We provide a detailed investigation of interaction effects, calendar and time-of-day effects, and industry-aggregation returns of various cross-sectional biases in the literature using a WLS Fama-Macbeth regression methodology on daily returns in the US equity markets from 1982 to 2011 and on intraday returns from 1993 to 2007. Among our findings regarding return effects are that 1) the reversal-momentum-reversal pattern in the short-, medium-, and long-term is highly variable by month, that 2) the industry momentum effect, as initially reported in Moskowitz and Grinblatt (1999) has largely disappeared according to the given methodology, and that 3) while intraday cross-sectional return variation displays periodicity effects as described by Heston, Korajczyk and Sadka (2010), the return structure varies significantly by time of day, unlike their report. Additionally, we also find that the “linearity” of a stock's past returns, as well as the skewness of the returns, have power in predicting the cross-section of stock returns; the results for skewness provide some empirical support for the results of Barberis and Huang (2008). For the size, value, risk, and turnover factors that we test, returns are generally much stronger in January than in other months, although industry aggregates general show little predictive power (with a few exceptions), echoing the results of Asness, Porter, and Stevens (2000). Finally, we implement a testing scheme that evaluates returns to portfolios that capture some of the pricing biases, taking into account various real-world constraints and trading costs. We find that 1) there are significant risk-adjusted returns to semi-active “structured” portfolios that arbitrage the noted biases (net of trading costs, given the constraints), especially after 2002, but that 2), using a short-scale time frame for calculating IR encourages benchmark hugging and suggests a semi-passive portfolio over active portfolios
