Introduction: This study explored social determinants of chronic child malnutrition on national level and in very poor ruralities in the Andean region in Peru. Methods: Two samples were studied, a national sample, (n = 1426) and a rural subsample, (n = 171). The samples consisted of women age 15-49 and their firstborn children 3-60 months of age. Data were provided by the Demographic and Health Survey ENDES Continua 2004-2006 (DHS 2004-2006). The variable for child malnutrition was composed by anthropometric measures of height and age, and 2 standard deviations below the WHO growth standard indicated chronic malnutrition or stunting. The influence of a range of variables on child stunting were tested through bivariate and logistic regression analyses performed in SPSS. Results: For the national sample, the traditional socioeconomic indicators were found to be significantly associated with child stunting. Especially wealth, education and occupation were strong predictors even after controlling for possible confounding variables like age, altitude, and ethnicity. For the rural subsample, none of the classical socioeconomic measures were statistically significantly related to child stunting, indicating that other factors play a greater role in reducing child malnutrition. Altitude was a strong predictor for both samples. Discussion. The diverging results comparing national and rural sample indicate that the widely used measures of income/wealth, education and occupation are not successful at explaining differences in health in very poor ruralities. Rather structural, natural, and cultural structures not addressed in the survey might play more important roles in producing differences in child growth. The results from this study support findings from other studies on social determinants of health in remote and extremely poor areas. Conclusion: The differing results between the national and rural sample show that one must be careful in generalising findings concerning determinants of health from national to regional levels.