We develop and estimate a panel data model explaining the answers to questions about subjective probabilities, using data from the US Health and Retirement Study. We explicitly account for nonresponse, rounding, and focal point “50 percent” answers. Our results indicate that for three of the four questions considered, almost all 50 percent answers can be explained by rounding. We also find observed and unobserved heterogeneity in the tendencies to report rounded values or a focal answer, explaining persistency in 50 percent-answers over time. Incorporating rounding and focal answers changes some of the conclusions about the socio-economic factors that determine expectations.