This report is written with the goal of analyzing the efficiency of the Hyndman-Ullah (HU) and weighted Hyndman-Ullah (wHU) methods when working with the populations that suffered higher mortalities due to the wars and pandemics. Accordingly, the HU and wHU methods are applied to the training sets containing outlying years. The historical data are obtained from Human Mortality Database, while forecasts and analysis are performed using R-package demography. Two countries, Sweden and France, are chosen to participate in the in-sample mortality forecast. By comparing forecasted mortalities for those countries after First World War (WWI), Spanish flu and Second World War (WWII), the forecast accuracy of both HU and wHU methods is evaluated. The wHU method proved better when working with the Swedish training set with moderate jumps in mortality. For the French data set consisting of large mortality jumps, both methods recorded significant error measures, which were decreased eventually by extending the training set. After that, the selection of four countries (Sweden, Denmark, Spain and Japan) provides the out-of-sample mortality forecast after the pandemics Covid-19 for the horizon of 30 years.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:mdh-58567 |
| Date | January 2022 |
| Creators | Raspudić, Teo |
| Publisher | Mälardalens universitet, Akademin för utbildning, kultur och kommunikation |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0017 seconds