Résumé en anglais

We use density forecasts derived from recursively estimated quantile regressions to calculate a forecast of the physical skewness of an asset's future return distribution. The forecast is unbiased and efficient, and it can easily be adapted to forecast the skewness of returns calculated over any conceivable return interval. Using Neuberger's (2012) realized physical skewness, we show that our quantile regression skewness forecast outperforms other variables proposed in the literature. Despite this, it does not condition the cross-section of future stock returns, neither independently nor when combined with other forecasts. Overall, we cast doubt on whether stock markets price expected stock skewness.