This research analyzes data collected by the Young Lives project on childhood poverty. It addresses the effect of socio-demographic, individual, and linguistic variables -- focusing on the latter -- on reading comprehension in Spanish by 502 Peruvian Quechua-Spanish bilingual children aged 8 years. The regression model tested explained 46% of the variance found: beyond the effect of non-linguistic factors, receptive vocabulary -- followed by listening comprehension and reading accuracy -- played the most important role. The importance of specifically promoting vocabulary development in the classroom over skills related to decoding is underlined.