Computerised Detection of Age, Period, and Cohort Effects

Abstract

Any mortality analysis is faced with the difficulty of dealing with large quantities of noisy data. Pragmatic approaches have ever relied on graphical analysis of mortality contour maps in order to identify age, period and cohort effects. We adapt and apply suitable methods of image processing and computer vision to demography. In particular, we design a multistage algorithm based on the well known Canny operator (Canny 1986) with a view to detecting abrupt changes in incremental mortality development factors by age over time. These edges indicate the boundaries between areas of higher and lower mortality improvements. The computerized detection algorithm attempts to increase the human ability to discover mortality effects. The approach further allows for a more objective judgement concerning the existence of a particular pattern in a given mortality surface. In particular, our aim is further to elucidate how computerized detection of age, period and cohort effects may complement demographic analyses by introducing a new technique of descriptive and exploratory data analysis.