Background: Evidence points to diverse risk factors associated with small- (SGA) and large-for-gestational-age (LGA) births. A more comprehensive understanding of these factors is imperative, especially in vulnerable populations. Objectives: To estimate the occurrence of and sociodemographic factors associated with SGA and LGA births in poor and extremely poor populations of Brazil. Methods: The study population consisted of women of reproductive age (14–49 y), whose last child was born between 2012 and 2015. INTERGROWTH 21st consortium criteria were used to classify weight for gestational age according to sex. Multinomial logistic regression modeling was performed to investigate associations of interest. Results: Of 5,521,517 live births analyzed, SGA and LGA corresponded to 7.8% and 17.1%, respectively. Multivariate analysis revealed greater odds of SGA in children born to women who selfreported as black (OR: 1.21; 95% CI: 1.19, 1.22), mixed-race (parda) (OR: 1.08; 95% CI: 1.07, 1.09), or indigenous (OR: 1.11; 95% CI: 1.06, 1.15), were unmarried (OR: 1.08; 95% CI: 1.07, 1.08), illiterate (OR: 1.47; 95% CI: 1.42, 1.52), did not receive prenatal care (OR: 1.57; 95% CI: 1.53, 1.60), or were aged 14–20 y (OR: 1.21; 95% CI: 1.20, 1.22) or 35–49 y (OR: 1.12; 95% CI: 1.10, 1.13). Considering LGA children, higher odds were found in infants born to women living in households with ≥3 inadequate housing conditions (OR: 1.11; 95% CI: 1.10, 1.12), in indigenous women (OR: 1.22; 95% CI: 1.19, 1.25), those who had 1–3 y of schooling (OR: 1.18; 95% CI: 1.17, 1.19), 1–3 prenatal visits (OR: 1.16; CI 95%: 1.14, 1.17), or were older (OR: 1.26; 95% CI: 1.25, 1.27). Conclusions: In poorer Brazilian populations, socioeconomic, racial, and maternal characteristics are consistently associated with the occurrence of SGA births, but remain less clearly linked to the occurrence of LGA births.