Empirical assessment of the applicability of shrinkage predictors in regression when past and future data differ