I am running probit for cross-sectional data that runs for several years (each individual still appears one time in the data). I want to see any changes in behavior after 2014. Is it better to run two models one for pre-2014 and one for post-2014 with year dummies, knowing that these are two different groups of people. Or, is it better to run one model only for all but with a dummy if the year is after-2014? any help is appreciated.