Objective The aim of this population-based validated study was to determine the course of tension-type headache and migraine and to evaluate the predictors of persistence. Methods We evaluated the course of headache in a large population from the first assessment in 2008 through a second assessment in 2013. Then we examined the factors associated with persistent migraine and persistent tension-type headache. Results Our study in 2013 revealed that only 42.9% of definite migraineurs in 2008 received the same diagnosis again, and of the remaining migraineurs 23.3% were newly diagnosed as definite tension-type headache; 11.6% evolved into probable tension-type headache, 6.4% changed to probable migraine, and 15.8% were headache free. The 17.7% of patients with definite tension-type headache in 2008 were newly diagnosed as having probable tension-type headache, 14.7% as having definite migraine, 6.4% as having probable migraine, and 28.9% as headache free in 2013, and only 32.3% received the definite tension-type headache diagnosis again. Binary logistic regression analysis showed nausea, throbbing and severe headache were the significant parameters for persistent migraine. A multiple regression analysis model with stepwise variable selection revealed that nausea, throbbing and severe headache and osmophobia remained in the final model as predictors of migraine persistence. We found no predictive factor for persistent tension-type headache. Conclusion Migraine and tension-type headache did not seem to show a simple bidirectional linear worsening from headache-free state to definite migraine or vice versa, hence the transitions between them are more chaotic, reflecting that there are still unknown modifiers and modulators. Certain headache characteristics of migraine might predict persistent migraine.