Does the Ross recovery theorem work empirically?

Starting with the fundamental relation that state prices are the product of physical probabilities and the stochastic discount factor, Ross (2015) shows that, given strong assumptions, knowing state prices suffices to back out physical probabilities and the stochastic discount factor at the same time. We find that such recovered physical distributions based on the S&P 500 index are incompatible with future returns and fail to predict future returns and realized variances. These negative results are even stronger when we add economically reasonable constraints. Simple benchmark methods based on a power utility agent or the historical return distribution cannot be rejected.