We estimate the effect of education on lifetime earnings by distinguishing between individuals who lived in rural or urban areas during childhood and between individuals with access to many or few books at home at age 10. We instrument years of education using compulsory school reforms and find that, whereas individuals in rural areas were most affected by the reforms, those with many books enjoyed substantially higher returns to their additional education. We show that books retain explanatory power even when we select relatively homogeneous groups in terms of the economic position of the household.