Brecha salarial de género en Chile: regresión lineal versus regresión simbólica.
Loading...
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
La brecha salarial de género se define como la diferencia de salario que recibe una persona con respecto a otra debido a su género, manteniendo constantes todas las variables. Es un tema ampliamente estudiado en la literatura y existe gran diversidad de resultados. En Chile, la literatura también presenta una gran diversidad de resultados dependiendo de los métodos empíricos utilizados.
La presente memoria de título estima la brecha salarial de género en Chile utilizando dos métodos de estimación. Por un lado, el método tradicional de la literatura de regresión lineal múltiple. Por otro lado, el método de regresión simbólica, un algoritmo que utiliza Machine Learning para estimar una ecuación en base a los datos.
Se utilizaron datos de la Encuesta de Caracterización Socioeconómica Nacional (CASEN) para 1992, 2000, 2011 y 2022. Para ambos métodos empíricos, se trabaja con una ecuación de salario a nivel individual donde la variable dependiente es el logaritmo natural del salario por hora del trabajador y el coeficiente clave es el parámetro que acompaña a la variable mujer (o sexo del trabajador).
Los resultados indican que el método que mejor ajusta los datos, considerando el error cuadrático medio (MSE), es la regresión lineal múltiple. La cual obtiene una disminución del 8,83% del MSE para 2022. La brecha salarial de género en Chile para 2011 y 2022 utilizando regresión lineal es de 16,29% y 11,76%, respectivamente. Mientras que utilizando regresión simbólica para 2011 es de 20,50% y para 2022 de 13,05%. Se aprecia que utilizando ambos métodos la brecha salarial de género disminuye en 2022 con respecto a las estimaciones para periodos anteriores. Dependiendo de la metodología, cambia la magnitud en la que disminuye la brecha salarial de género, obteniendo una mayor disminución cuando se utiliza regresión simbólica.
Se concluye finalmente que el método de regresión lineal permite mejorar la precisión de la estimación de la brecha salarial de género. Es importante destacar que los resultados respecto a esta brecha son distintos, pero cercanos en magnitud. Al utilizar regresión simbólica, la brecha salarial tiene mayores valores lo que puede tener impacto para el diseño de políticas.
The gender wage gap is defined as the difference in salary that a person receives with respect to another due to their gender, keeping all variables constant. It is a widely studied topic in literature and there is a great diversity of results. In Chile, the literature also presents a great diversity of results depending on the empirical methods used. This Tesis estimates the gender wage gap in Chile using two estimation methods. On one hand, the traditional method of multiple linear regression literature. On the other hand, the symbolic regression method, an algorithm that uses Machine Learning to estimate an equation based on data. Data from the National Socioeconomic Characterization Survey (CASEN) were used for 1992, 2000, 2011 and 2022. For both empirical methods, a wage equation at the individual level is used, where the dependent variable is the natural logarithm of the worker’s hourly salary, and the key coefficient is the parameter that accompanies the woman variable (or worker’s sex). The results indicate that the method that best fits the data, considering the Mean Square Error (MSE), is symbolic regression, obtaining an 8,83% reduction in MSE for 2022. The gender wage gap in Chile for 2011 and 2022 using multiple linear regression is 13,29% and 11,76%, respectively. Meanwhile, using symbolic regression, it is 20,50% for 2011 and 13,05% for 2022. Using both methods, the gender wage gap decreases in 2022 compared to estimates for previous periods. Depending on the methodology, the magnitude of the gender wage gap decrease varies, obtaining a greater decrease when using symbolic regression. Finally, it is concluded that the multiple linear regression method improves the accuracy of estimation of the gender wage gap. It is important to note that results regarding this gap are different, but similar in magnitude. When using symbolic regression, the gender wage gap has larger values, which can impact policy design.
The gender wage gap is defined as the difference in salary that a person receives with respect to another due to their gender, keeping all variables constant. It is a widely studied topic in literature and there is a great diversity of results. In Chile, the literature also presents a great diversity of results depending on the empirical methods used. This Tesis estimates the gender wage gap in Chile using two estimation methods. On one hand, the traditional method of multiple linear regression literature. On the other hand, the symbolic regression method, an algorithm that uses Machine Learning to estimate an equation based on data. Data from the National Socioeconomic Characterization Survey (CASEN) were used for 1992, 2000, 2011 and 2022. For both empirical methods, a wage equation at the individual level is used, where the dependent variable is the natural logarithm of the worker’s hourly salary, and the key coefficient is the parameter that accompanies the woman variable (or worker’s sex). The results indicate that the method that best fits the data, considering the Mean Square Error (MSE), is symbolic regression, obtaining an 8,83% reduction in MSE for 2022. The gender wage gap in Chile for 2011 and 2022 using multiple linear regression is 13,29% and 11,76%, respectively. Meanwhile, using symbolic regression, it is 20,50% for 2011 and 13,05% for 2022. Using both methods, the gender wage gap decreases in 2022 compared to estimates for previous periods. Depending on the methodology, the magnitude of the gender wage gap decrease varies, obtaining a greater decrease when using symbolic regression. Finally, it is concluded that the multiple linear regression method improves the accuracy of estimation of the gender wage gap. It is important to note that results regarding this gap are different, but similar in magnitude. When using symbolic regression, the gender wage gap has larger values, which can impact policy design.
Description
Tesis presentada para optar al título de Ingeniero/a Civil Industrial.
Keywords
Género, Desigualdades económicas, Interpretación estadística de datos