Abstract
We propose a detailed and comprehensive examination of the two main regression-based techniques used to detect herding among investors. We also introduce a novel approach based on the autocorrelation of returns. We test all models on a unique dataset of wine prices. For the first two models, our conclusions highlight the importance of macroeconomic variables (US equities) on the dispersion of wine returns. Thus, if wine investors herd, it is essentially because of external contingencies and they are not driven by the state of the wine market itself. The third (new) model seems to indicate that there is at most weak evidence of herding and the conclusions are robust when controlling for the state of the US equity market.