Andy Chen on cross-sectional asset pricing
From Andrew Chen: "Is Everything I was Taught About Cross-Sectional Asset Pricing Wrong?!" | RR 316 · · TheRationalReminderPodcast
“There are basically three explanations for why cross-sectional predictors exist: risk (they compensate you for bearing risk), mispricing (prices aren't right), or the statistics being wrong somehow.”
On , Andy Chen, Vice President & Treasurer at General Dynamics Corp, spoke about cross-sectional asset pricing during Andrew Chen: "Is Everything I was Taught About Cross-Sectional Asset Pricing Wrong?!" | RR 316 on TheRationalReminderPodcast.
About Andy Chen
Andrew Chen, a principal economist at the Federal Reserve Board, appeared on the Rational Reminder podcast in August 2024 to discuss his research on cross-sectional asset pricing. Chen stated that his views are his own and not necessarily those of the Federal Reserve. He described a predictor as a variable that forecasts asset returns, which can be turned into a factor or trading strategy. Chen outlined three explanations for why cross-sectional predictors exist: they compensate for risk, they result from mispricing, or the statistics are wrong. He noted that academic finance faced a trade-off between open collaboration and close competition, leading his team to create an open-source dataset. Chen reported that of roughly 300 variables collected, about 200 predicted returns in original papers, and his replication judgment replicated all but three, a failure rate of roughly 1–2%. Chen discussed the decay of anomaly returns, stating that if a paper's sample ends in 1989, anomaly returns decline by about 50% from 1990 onward. He said transaction costs eat up about 25–30% of trading-strategy returns in original samples. Chen observed a kink around the mid-2000s in factor returns, which he hypothesized was due to the internet making information more accessible. He suggested that data-mining accounting ratios in 1980 would have uncovered anomalies decades before publication. Chen advised trusting numbers more than text in academic papers, verifying that documented data supports claimed mechanisms.