Efectos del endeudamiento de los hogares mexicanos en su ahorro y consumo: Un enfoque de Ciencia de datos

Effects of Mexican Household Indebtedness on Their Savings and Consumption: A Data Science ApproachThis research aims to group samples of indebted Mexican households that share similar socioeconomic attributes using the k-means algorithm so that nonlinear models are estimated to measure the effects of each group's debt on their savings and consumption. The algorithm was implemented on indebted households included in the ENIGH 2018. As a result, four clústers were formed where one stood out by making up 3.4% of the sample; however, its average indebtedness rate exceeds the average indebtedness rate by 53 percentage points from the rest of the clústers. Modern clústering techniques are recommended to utilize the abundance of official data and develop data-driven economic policies targeted at particular population groups. The originality of this research is based on the use of an unsupervised algorithm for the choice of the studied sample. In conclusion, the households with the highest levels of over-indebtedness are made up of those where the head has higher education, regardless of the income decile to which the household belongs. ; El objetivo de esta investigación es agrupar muestras de hogares endeudados mexicanos que compartan atributos socioeconómicos similares mediante el algoritmo k-medias de manera que se estimen modelos no lineales para medir los efectos de la deuda de cada grupo en su ahorro y consumo. El algoritmo se implementó sobre hogares endeudados incluidos en la ENIGH 2018. Como resultado se conformaron cuatro clústeres donde uno sobresalió al conformar el 3.4% de la muestra, sin embargo, su tasa de endeudamiento promedio excede en 53 puntos porcentuales a la tasa de endeudamiento promedio del resto de los clústeres. Se recomienda el uso de técnicas de agrupación modernas para aprovechar la abundancia de datos oficiales y para elaborar políticas económicas, basadas en datos, dirigidas a grupos particulares de la población. La originalidad de esta investigación se basa en el uso de un algoritmo no ...