[ES] Este estudio permitió consolidar el conocimiento sobre los determinantes del consumo de hojas de coca en Bolivia. Estimando modelos pooled condicionales y aplicando técnicas de Machine Learning utilizamos las Encuestas de Hogares disponibles en el INE de Bolivia. Existen dos poblaciones en riesgo de convertirse en un problema de bienestar estructural de largo plazo, las naciones y pueblos originarios campesinos, y aquellos que están en extrema pobreza. Aparece el efecto compensatorio sustituyendo bienes de la canasta básica por hojas de coca y desaparece a mayor grado educativo e ingreso. El fumar y beber alcohol se comportan como bienes complementarios. [EN] This study consolidates knowledge about the determinants of coca leaf consumption in Bolivia. Applying conditional pooled models combined with Machine Learning techniques, use the Household Surveys available at the Bolivian INE. There are two populations at risk of becoming a long-term structural well-being problem indigenous peasant native nations and peoples, and those who are in extreme poverty. The compensatory effect appears by substituting goods from the basic food basket for coca leaves, which disappears when they have higher educational levels and better income. Smoking and drinking alcohol behave as complementary goods.