Comparación de Ponderaciones en Regresiones Probit Simultáneas en un Modelo para la Estimación de la Participación Laboral

Estimations of labor participation, based on samples of complex design, generally do not discuss the use of weights; implicitly considering that the data came from a simple random sampling. This paper studies the effects of the application of three weighing alternatives : i) not to weigh ii) to assign weights which are proportional to agglomerate size iii) to select a sub-sample which allow self-representation. The application is carried out for a wave of the Permanent Household Survey (Encuesta Permanente de Hogares - EPH) on a model of simultaneous probit equations. The evaluation of the estimates obtained suggests that not to weigh produces no significant biases and represents an antecedent to the empirical estimates with nonlinear models for a system of equations with limited dependent variables.