The presidential primary elections would be an interesting topic to try to apply survival analysis to. For example, the democratic primary has a large # of candidates running in their primary in 2019, and the previous republican primary had a large # of candidates: therefore, there would be enough data for analysis. Most political analysis covering the elections tracks poll numbers and assumes this is a good indication of performance, but there is generally no modeling or attempt to estimate formally which candidates are most likely to drop out.

Suppose you want to study which factors lead to faster drop-outs. For this study, what's your time to event? Should it be measured in days, months or years?