Abstract Herding behavior has been in the epicenter of a heated debate over the past three decades across traditional and alternative asset markets. Establishing the existence of herding relies on a standard regression-based testing procedure. Employing Monte Carlo simulations we show that spurious anti-herding behavior might emerge even if the series are random and totally uncorrelated, provided that the residuals of the model fail to conform to some of the assumptions of standard linear regression. The simulations from a t-student distribution are examined as a function of the degrees of freedom, the length of the simulations, and the number of series.