Document type
Bachelor thesisPublication date
Publication license
Please use this identifier to cite or link to this item: https://hdl.handle.net/2445/126082
Predicció de la taxa d’atur espanyola: un anàlisi regional
Journal Title
Authors
Director/Tutor
Journal ISSN
Volume Title
Related resource
Abstract
(cat) Aquest projecte d’investigació té per objectiu principal la realització i estudi previ de les prediccions de l’atur espanyol, posant èmfasi a l’anàlisi regional. Per un costat, es fa ús dels mètodes deterministes per la taxa de desocupació nacional. Mitjançant els contrastos de Daniel i Kruskal-Wallis es detecta que la sèrie temporal presenta tendència però no component estacional. És per això que els mètodes emprats en la primera fase del exercici són: la TL (Tendència Lineal), les DMM (Dobles Mitjanes Mòbils) i l’AEH (Allisat Exponencial de Holt). Per l’altre costat, es generen prediccions a través de models ARIMA (Model Autoregressiu de Mitjana Mòbil) a nivell nacional i per a cada una de les disset CCAA (Comunitats Autònomes). En darrer lloc, s’avalua la capacitat predictiva dels mètodes i models utilitzats. La conclusió principal d’aquest estudi subratlla la importància de generar prediccions independents per a cada regió.
(eng) The goal of this research project is to conduct a forecasting experiment for the Spanish unemployment rate at a regional level. We use a sampling size that starts at the first quarter of the year 2002 and ends at the last quarter of the year 2017. First, we apply a deterministic approach at the national level. We run Daniel and Kruskal-Wallis tests and find that the unemployment growth rate presents a significant trend component although the seasonal component is not significant. As a result, we use the Linear Trend, Double Moving Average and Holt Exponential Smoothing methods. Second, we generate forecasts by means of ARIMA models at the national and regional level. At the national exercise we use the Box-Jenkins Methodology with Gretl and compare it to the output generated with the auto-ARIMA function with R designed by Hyndman and Khandakar (2008). We find that the auto-ARIMA function provides better outcomes, for this reason we use it for model selection at the regional level. We obtain different optimal ARIMA models in each region. This finding suggests that unemployment forecasting should be implemented by means of region-specific models. Finally, we compute the out-of-sample forecast accuracy for the four quarters of the year 2017. We obtain the best results for Galicia and Valencian Community, as opposed to Aragon.
(eng) The goal of this research project is to conduct a forecasting experiment for the Spanish unemployment rate at a regional level. We use a sampling size that starts at the first quarter of the year 2002 and ends at the last quarter of the year 2017. First, we apply a deterministic approach at the national level. We run Daniel and Kruskal-Wallis tests and find that the unemployment growth rate presents a significant trend component although the seasonal component is not significant. As a result, we use the Linear Trend, Double Moving Average and Holt Exponential Smoothing methods. Second, we generate forecasts by means of ARIMA models at the national and regional level. At the national exercise we use the Box-Jenkins Methodology with Gretl and compare it to the output generated with the auto-ARIMA function with R designed by Hyndman and Khandakar (2008). We find that the auto-ARIMA function provides better outcomes, for this reason we use it for model selection at the regional level. We obtain different optimal ARIMA models in each region. This finding suggests that unemployment forecasting should be implemented by means of region-specific models. Finally, we compute the out-of-sample forecast accuracy for the four quarters of the year 2017. We obtain the best results for Galicia and Valencian Community, as opposed to Aragon.
Description
Treballs Finals del Grau de d'Administració i Direcció d'Empreses, Facultat d'Economia i Empresa, Universitat de Barcelona, Curs: 2017-2018 , Tutor: Oscar Claveria González
Subject (English)
Citation
Citation
FIGUEROLA WISCHKE, Anton. Predicció de la taxa d’atur espanyola: un anàlisi regional. [consulted: 8 of June of 2026]. Available at: https://hdl.handle.net/2445/126082