The logit model of educational transitions has become standard in research
in educational stratification. One limitation of the model, however is the
assumption that individuals progress through the educational system in a un
ilinear sequential mode. Many school systems contain parallel branches of s
tudy that are most fruitfully seen as qualitatively different alternative p
athways with different probabilities of school continuation attached to the
m. This study tests a multinomial model of educational careers, that takes
previous paths and grade-point averages into account. Applied to a large Sw
edish longitudinal data set the model rests whether conclusions about class
stratification in educational attainment based on a logit model are borne
out. Results show that the pathway a student has taken through the school s
ystem influences the probability of making subsequent educational transitio
ns. This result is robust to unmeasured heterogeneity modeled using a laten
t class approach. Furthermore, the traditional logit model tends to deflate
class-origin effects at early transition points while inflating them at th
e transition to higher education. The results give some support to the hypo
thesis that origin effects are strongest at more "indirecr" and unusual pat
hways.