The large gap in college-going by family income is caused by two reasons: liquidity constraints and low return to schooling among the poor. We use scores of the matriculation exams which are held at the end of secondary school in Israel ('Bagrut exams') to separate between the two reasons. We find that after controlling for the matriculation scores the return to college among male students is unrelated to the family income. However, among female students the return to college is larger among students from poor families (after controlling for the exam scores). Hence, the gap in college-going by family income among male which does not stem from the gap in matriculation scores is caused by liquidity constraints. Among female the corresponding gap (that which does' not stem from the matriculation scores) underestimates liquidity constraints.