Recent actual experiment (being written up by a student as a science fair project):

Obtained the actual birth date, place and time for 35 students, grades 6 through 8.

Got detailed personality profiles from a source recommended by a true believer in astrology. For each student, got a "correct" one, and three others -- wrong date, wrong place, wrong gender (with all other data correct).

Each student was given their four profiles, and spent a week pondering them, showing them to friends and family, and deciding which was the "best fit".

Final score: out of 35 students, 7 picked the "correct" profile. Darn close to the 25% expected for random chance.

Best/most revealing student comment: as we discussed the results, one of the students said "I guess we just aren't very good at astrology" - implying that the astrology must be correct, they just weren't able to "read" it properly!