Important note: We have issued CMI_2017 v4 in order to correct an issue with CMI_2017 v3.
The issue in CMI_2017 v3 is that if a user:
- selects the “Extended” option for initial improvements; and
- selects a value for the period smoothing parameter, Sk, which is not one of the pre-calculated values (6.5, 6.75, 7, 7.25, 7.5, 7.75, 8, 8.25 and 8.5) then the software will not apply the correct smoothing to the initial mortality improvements.
The issue only affects CMI_2017 v3. It does not affect CMI_2017 v1, CMI_2017 v2, or versions of the software for other years.
Please note that we became aware in 2020 that the calculation of deviance and deviance residuals in CMI_2017 is incorrect when age/year cells in the calibration datasets have no deaths. This does not affect the Core dataset, and only has a marginal impact on the fitted model for other datasets that we have tested. Because of the limited impact, and the time since the release of CMI_2017, we have not issued a corrected version of the Model.
An updated version of the CMI_2017 software, v02, was released in August 2018 to allow the "live link" to continue to work following changes to the Human Mortality Database website. A further version, v03, was released in December 2018 to allow the smoothing parameters to be adjusted for populations of different sizes, as described in Working Paper 115. Neither change affects the CMI_2017 projections.
A Briefing note on CMI_2017 has also been issued and is publicly available alongside this paper. It provides an overview of the latest version of the Model and is aimed at those receiving advice based on the Model, such as pension scheme trustees and Non-Executive Directors of insurance companies.
The latest version of the Model is calibrated to England & Wales population mortality data up to 31 December 2017. Its method is the same as CMI_2016 apart from using an updated dataset, and a minor change to the method used to construct the dataset.
Core Projections generated by CMI_2017 produce lower expectations of life than those produced by CMI_2016 (with other parameters held constant) for all ages for both males and females.
Working Paper 105 contains numerical results from CMI_2017, and commentary. It also contains analysis of recent mortality in the general population and illustrations of the sensitivity of the Model to new data.
The Model is accompanied by a methods paper, containing technical details of the calculations, and a user guide, describing how to use the software. These are updated versions of Working Papers 98 and 99. Two additional spreadsheets have been released to show the derivation of the CMI_2017 dataset.
The software has been tested and found to work properly using Excel 2010 and Excel 2016 in a Windows environment. If you see a yellow banner in Excel with buttons saying e.g. ‘Enable Editing’ or ‘Enable Content’ then you will need to click on these buttons in order to run the software. If however you see a red banner and the button to proceed is disabled preventing you from running the software than please let us know at Projections@cmilimited.co.uk.
The Committee hosted a meeting at Staple Inn on 3 April 2018, in conjunction with the Staple Inn Actuarial Society (SIAS), to discuss recent mortality experience and the prospects for future mortality improvements. The presentation is available from the SIAS website.
Documents
- CMI Working Paper 105
- Briefing note for Working Paper 105
- CMI_2017 methods
- CMI_2017 software user guide
- The updated Model, CMI_2017
- Derivation of the male dataset used in CMI_2017
- Derivation of the female dataset used in CMI_2017
View the previous versions of the CMI Mortality Projections Model
Events calendar
No results found.