Cuantificación del riesgo global del asegurado para mejorar la tarificación
- Autores: Alemar E. Padilla Barreto
- Directores de la Tesis: Montserrat Guillén Estany (dir. tes.), Catalina Bolancé Losilla (codir. tes.), Josep Ginebra (tut. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2019
- Idioma: español
- Tribunal Calificador de la Tesis: José María Sarabia Alegría (presid.), Pedro Delicado (secret.), Isabel Serra Mochales (voc.)
- Programa de doctorado: Programa de Doctorado en Estadística e Investigación Operativa por la Universidad Politécnica de Catalunya
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
The first contribution of this Thesis is the creation of a new risk measure that allows to classify the clients according to the risk that they bring to the company. This measure defined from the aggregate information at the client level is easy to understand, implement and also takes into account the propensity to renew, the claims and the relationship between the different policies subscribed by the insured.
Subsequently we focus on one of the factors that affect the risk measure defined previously, specifically in the analysis of the propensity to renewal, which is the second contribution of this work. First, from an univariate point of view and for two different lines of business, which are assumed independent of each other - Home and Motor -, the predictive capacity of different alternative churn prediction models is evaluated according to different adjustment criteria ("churn models" are those that allow evaluating the probability that clients cancel their policies and leave the company). A comparative analysis is established between a classic univariate regression model and two machine learning models. The adjustment criteria proposed in this part represent a simple and practical tool in order to choose the threshold from which it is possible to classify clients according to their propensity for renewal.
The joint modeling for customer lapses (in more than one type of policy) can be considered as our third contribution. In this part we analyze the propensity to renew the policies of customers with two types of insured risks simultaneously. That is, we take into account the existence of dependence between lines of business and show how the existence of dependence affects decisions about renewal in both types of policy - Home and Motor.
Our fourth contribution raises several reflections about the change that supposes for the insurance companies the digitalization of the information. In addition, we propose a practical application where, from the use of multinomial models, we obtain the simultaneous propensity to renew or not the policies that the same client has underwritten in the same company.
The following contributions correspond to the second part of the Thesis and are directly related to the risk quantification in finance and insurance. In this sense, a first contribution consists of analyzing the effect that the selection of the dependency model has on the estimation of value at risk. In the same vein, the second contribution in this part analyzes the weaknesses and strengths of nonparametric methods in the quantification of risk.
