Methods for disentangling period and cohort changes in mortality risk over the twentieth century: comparing graphical and modelling approaches

This paper explores changes in age--specific mortality risk across periods and cohorts during the twentieth century in the developed world. We use and compare two approaches---one graphical (Lexis plots) and one statistical (an adapted Hierarchical age--period--cohort model)---that control out overall trends in mortality, to focus on discrete changes associated with specific events. Our analyses point to a number of key global and local events in the Twentieth Century associated with period and/or cohort effects, including the World Wars and the influenza pandemic of 1918--19. We focus particularly on the UK but look at other countries where results are particularly noteworthy, either substantively or methodologically. We also find a decline in mortality in many western countries, specifically in the 1948 birth cohort, which may be associated with the development of post--war social welfare policies, the economic investment in Europe by the United States, the accessibility of antibiotics such as penicillin, and, in the UK, the founding of the NHS.
We finish by considering the advantages and disadvantages of using the two methods with different sorts of data and research questions.