CMI mortality projections model
\si-ɛm-aɪ\ \mɔrˈtæləti\ \prəˈʤɛkʃənz\ \ˈmɑdəl\
A deterministic model developed in the UK by the Continuous Mortality Investigation for projecting mortality rates into the future. (Often abbreviated to “the CMI model”).
The core model is calibrated in the short term to recent mortality experience observed in the England and Wales population (using P-spline techniques). These short-term mortality rates are then blended to a long-term rate selected by the user as a sustainable rate of improvements. Adjustable parameters can be used to refine the shape of the model in the medium term during the blending period. The long-term rate itself is tapered away between the ages of 90 and 120. Age-period effects and cohort effects on improvements are projected separately.
The first version of the model was released in 2009, with updated versions (mainly reflecting updates to the calibration data) released annually since. A more significant update to the model was made in 2016 with more advanced parameters being introduced. Most notably, the 2016 update introduced a period smoothing parameter (Sĸ), for the user to adjust how much credibility to place on recent data versus how much on more historical data in the model’s calibration. Also, the core period of tapering away of the long-term rate was changed to occur between the ages of 85 and 110. A further advanced parameter, initial addition to mortality improvements parameter (A), was introduced in the 2018 model, which allows users to alter the short-term rates.
The model is released with a set of ‘core parameters’ which are designed to reflect a reasonable improvement scale for the population of England and Wales. Adjustments may be required to the core parameters if the population being modeled is experiencing significantly different improvement effects compared to the population of England and Wales as a whole.