Improving inference using probabilistic expectations

The focus of this paper is on the use of probabilistic expectations in economic modeling and the effect of rounding on such modeling. A bootstrap algorithm is proposed to obtain empirical parameter estimate bounds, confidence intervals and diagnostic distributions for partially identified models. The algorithm leads to stricter parameter bounds and confidence intervals than those obtained through standard estimation methods. Using the results from the algorithm the effect of focal point answers and probability weighting are assessed for predicting wealth based on the subjective survival probability. The analysis showed that deleting respondents who gave focal point answers did not improve inference and the results for using probability weighting were inconclusive.