Can personality explain the educational gradient in divorce? Evidence from a nationally representative panel survey