Forensic economists debate whether it is most appropriate to estimate future net discount rates using either historical averages or the current rate. We develop an efficient estimator based on the time series properties of the net discount rate and the horizon over which net discount rates are to be forecasted. The efficient estimator optimally combines the two standard approaches. Benchmark comparisons suggest that our estimator is at least twice as efficient as either the historical average or the current value. Empirical results from U.S. and international data support this finding. We also show that a very simple compromise estimator that equally weights the historical and current rates outperforms the standard estimators and is nearly as efficient as the optimal estimator.