**Background:** Early childhood development (ECD) is a multisectoral concern, particularly in contexts of poverty. Approximately 249 million children in lower- and middle-income countries may not develop fully due to stunting and poverty. In the Dominican Republic, 49.3% of children lived in poverty in 2012, and 64% of children aged 1–14 years experienced psychological aggression or physical punishment in 2019. While the association between poverty and developmental delay is established, the role of family moderating factors that may buffer these effects is unknown.
**Methods:** The authors conducted a secondary analysis of the Dominican Republic's 2014 MICS5 (n=7,356 children aged 36–59 months) and 2019 MICS6 (n=3,399 children aged 36–59 months) surveys. Four study aims were tested: 1) confirm the relationship between socioeconomic position (SP), parenting practices, and ECD using Pearson's correlations; 2) determine if a sociodemographic model (child's age, sex, family SP, mother's education) predicted ECD via hierarchical multiple regression; 3) determine if a psychosocial model (stimulating activities, homemade and store-bought toys, number of books, early childhood program attendance, positive and negative discipline) predicted ECD above and beyond the sociodemographic model; and 4) explore mothers' beliefs about physical punishment and its relationship with ECD using ANCOVA controlling for SP. The ECD index (range 0–1) had poor internal consistency (α=0.5).
**Key Results:** The 2019 sample had a slightly higher mean ECD score (0.68) than 2014 (0.67), with a negligible effect size (d=0.06). For Aim 1, SP correlated positively with ECD in both years (r=0.29 in 2014, r=0.23 in 2019; z=3.10, p=0.001). SP also correlated positively with stimulating activities, store-bought toys, number of books, early childhood program attendance, and positive discipline in both years. For Aim 2, the sociodemographic model significantly predicted ECD, accounting for 14.0% of variance in 2014 and 12.4% in 2019. The strongest sociodemographic predictors were SP (β=0.216 in 2014, β=0.186 in 2019) and mother's education (β=0.139 in 2014, β=0.141 in 2019). For Aim 3, adding psychosocial variables significantly increased variance explained by 6.3% in 2014 and 4.4% in 2019 (total R²=0.204 in 2014, 0.168 in 2019). The strongest psychosocial predictors in 2014 were: attendance at early childhood program (β=0.155), negative discipline (β=-0.126), number of books (β=0.122), and store-bought toys (β=0.080). In 2019: negative discipline (β=-0.133), number of books (β=0.106), stimulating activities (β=0.074), and early childhood program attendance (β=0.059). For Aim 4, children of caregivers who did not agree with physical punishment had significantly higher ECD scores (2014: F(1,7287)=25.28, p<0.001; 2019: F(1,2055)=14.62, p<0.001) and more stimulating activities at home, controlling for SP. Caregivers who agreed with physical punishment used both more positive and more negative discipline.
**Clinical Implications:** The findings confirm that both sociodemographic and psychosocial factors independently predict ECD in the Dominican Republic, with psychosocial factors explaining additional variance beyond sociodemographics alone. Negative discipline was the strongest negative predictor, while early childhood program attendance, number of children's books, and stimulating activities were the strongest positive predictors. These results support interventions that promote positive parenting, reduce physical punishment, increase access to books and stimulating activities, and expand early childhood education—particularly for children in poverty. However, the authors emphasize that while psychosocial interventions can buffer the effects of poverty, sustainable large-scale efforts must address the underlying structural problem of poverty itself to prevent developmental delays and achieve health equity. Limitations include the ECD instrument's poor internal consistency (α=0.5), the inability to include nutrition variables (only available for children 0–23 months), and the model explaining only 16.8–20.4% of ECD variance.