Abstract
Prospect Theory is well understood in contexts without a time dimension. In intertemporal contexts, however, it is unclear how prospect theory should be applied. In particular, it is unclear whether probabilities should be weighted within time periods or whether probabilities of present values should be weighted. Furthermore, it is unclear what parametric specifications of probability-weighting and value functions should be used. We find in a pre-registered experiment on a representative sample that the version weighting probabilities of present values predicts decisions best. Estimated probability weighting functions are inverse-S shaped, and value functions are almost linear.