Welcome to the UPF Digital Repository

Banking crisis prediction: exploring the role of text-based indicators

Show simple item record

dc.contributor.author Hunter, Timothy
dc.contributor.author Garcia Aguilar, Ferran
dc.date.accessioned 2023-12-05T15:12:40Z
dc.date.available 2023-12-05T15:12:40Z
dc.date.issued 2023-06-12
dc.identifier.uri http://hdl.handle.net/10230/58460
dc.description Treball fi de màster de: Master's Degree in Specialized Economic Analysis: Macroeconomic Policy and Financial Markets Program. Curs 2022-2023
dc.description Tutors: Luca Gambetti i Hugo Rodriguez
dc.description.abstract We study text indicators to improve banking crisis forecasts. Using newspaper headlines from 18 countries over 1950-2019, we apply latent dirichlet allocation (lDA) to extract topic frequencies summarising the economic news written about a country. Including these variables in a logit model, we find significant improvement in predicting banking crises, but there is evidence that this is due to random chance, and results are not replicated using random forests. This runs counter to findings from other authors, suggesting that lDA may be over-sensitive to specification. Thus, further research should refine the lDA process, to establish a standardised approach across research.
dc.description.abstract Estudiamos indicadores basados en texto para mejorar la predicción de crisis bancarias. Usando titulares periodísticos de 18 países durante 1950-2019, aplicamos la asignación latente de Dirichlet (lDA) para extraer frecuencias sobre temas latentes en las noticias. Incluyéndolas en un modelo logit, encontramos una mejora significativa en el rendimiento, pero hay indicios de que puede ser debido al azar. Además, los resultados no se replican utilizando random forests. Al contradecir los resultados de otros autores, sugerimos que lDA puede ser sensible a su especificación. En futuras investigaciones se debería estandarizar el proceso lDA para establecer un marco común en éstas.
dc.format.mimetype application/pdf
dc.language.iso eng
dc.rights This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License
dc.rights.uri https://creativecommons.org/licenses/by-nc-nd/4.0
dc.subject.other Treball de fi de màster – Curs 2022-2023
dc.title Banking crisis prediction: exploring the role of text-based indicators
dc.type info:eu-repo/semantics/masterThesis
dc.subject.keyword Banking crisis prediction
dc.subject.keyword Natural language processing
dc.subject.keyword Latent Dirichlet allocation
dc.subject.keyword Predicción de crisis bancarias
dc.subject.keyword Procesamiento de lenguajes naturales
dc.subject.keyword Asignación latente de Dirichlet
dc.rights.accessRights info:eu-repo/semantics/openAccess

Thumbnail

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account

Statistics

In collaboration with Compliant to Partaking