betareg on skewed proportions leads to a high predicted mean

Question

I\'m running a beta regression on proportions which are highly skewed (median ~2%, mean ~13%). However, the mean predicted value coming out of the model is typically about