How do quants know when a momentum trade is about to crash? The trigger is crowding, and it is invisible on any data feed. In 2022 Dong Lou and Christopher Polk built a way to read it: comomentum. Take the winner and loser stocks a momentum strategy holds, strip out the three Fama-French factors and the industry effect, and measure how correlated the leftover returns are. When the same crowd piles into the same names, they move together beyond their fundamentals. From 1965 to 2015 that abnormal correlation averaged 0.092, from 0.037 when quiet to 0.241 when packed. When comomentum sat in its top 20%, momentum returns ran about 12.7% lower the next year and another 13% lower the year after versus low-crowding periods, more than 25 points over two years (t = 3.35), and the share of days below -1% rose from 8.4% to 22.5%. Limitations: it only holds for high institutional-ownership stocks in the modern period, it is a one-to-two year effect rather than a short-run timing signal, and the authors frame it as crowded-trading dynamics, not proof of market inefficiency. Paper: Lou & Polk, Comomentum, Review of Financial Studies 35(7), 2022. DOI 10.1093/rfs/hhab117 #finance #quant #trading #algotrading #stocks
How do quants know when a momentum trade is about to crash? Lou and Pog found something in 2022. Let's talk about it. You take the winners and los...
