The Case of Herding ist Stronger than You Think

In case of herding, investors follow each other, prices move together more than they normally do, and the cross-sectional dispersion of returns decreases. Chang, Cheng, and Khorana (2000) suggest to test for herding by regressing the cross-sectional absolute deviation on the absolute and squared excess market return. They argue that there is evidence for herding in case of large market movements when the coeffcient of the squared excess market return is signifcantly smaller than zero. We show that the true coeffcient of the squared excess market return is positive under the null hypothesis of no herding. The test of Chang, Cheng and Khorana is thus biased against finding evidence for herding. We find that this bias matters. For the S&P 500, the test of Chang, Cheng and Khorana signals that there is no herding over the period from 2008 to 2013, while the modified test based on the correct null hypothesis provides clear evidence for herding.