Forecasting Philippine mortality rates using the Hyndman-Ullah methodology

• Main Authors: Hung, Steve Kristoffer G., Sia, James Oliver C.
• Format: text
• Language: English
• Published: Animo Repository 2014
• Subjects: Physical Sciences and Mathematics
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/17995
• Institution: De La Salle University

Mortality is a key factor in influencing several of a countrys sectors such as economic, social security, and insurance industries. However, due to changes in the human lifestyle brought about by factors such as advancements in technology, mortality has not remained constant through time. From this fact arose the need for regular accurate mortality modeling and forecasting. Studies on Philippine mortality have often been difficult due to the unavailability and inadequacy of data. This paper proposes the use of a nonparametric and functional data approach on 1980- 2009 Philippine death rates, as proposed by R. J. Hyndman and M. S. Ullah in 2007. Through this robust method, mortality data is first smoothed from noise, consequently addressing the issue regarding data grouped in age intervals. Functional principal component analysis is performed, fitting a basis expansion model to the smoothed data. Using a robust ARIMA method, the model coefficients are then forecasted which lead to mortality rate forecasts. With these, more comprehensive analysis and interpretation of mortality behaviors over age and time are achieved.