What age definition is used for mortality rates in the CMI tables?

The CMI uses the standard definition of q(x) in its tables; that is q(x) is the probability that a life aged exactly x will die in the next year ie, before attaining exact age x+1. Similarly, q(x+.5) is applicable from exact age x+.5 to exact age x+1.5. This notation is described in most standard texts on the subject, eg, page 8 of Life Contingencies by Alistair Neill. For the avoidance of doubt, the age definition in the tables is not next, nearest or last but exact.

Where do I find published mortality tables?

  • The 00 series base mortality tables were published in CMI Working papers 21 and 22.
  • In 2002, an interim basis for adjusting the 92 series mortality projections was published in CMI Working paper 1
  • The 92 series base mortality tables and projections were published in CMI Report 17.
  • The 80 series base mortality tables and projections were published in CMI Report 10.
  • Earlier mortality tables were published in various printed volumes which are not available electronically.
  • All published CMI base mortality tables are available in Excel format. 

How are joint life cases treated?

Joint life cases are treated as a single life that is in force in both the male and female investigations. When the first life dies it is treated as a death in that investigation. The second life then exits from the other investigation. The exposed to risks for the two lives are the same except in the last year where the death is exposed for a full year and the other life is exposed for half a year.

The 00 series tables are described as giving mortality rates applicable to the calendar year 2000 - but at what exact point in 2000 do they apply?

For the 00 series tables, the value of qx applies on average to a life attaining age x in the middle (i.e. midnight between 30 June and 1 July) of 2000 and gives the probability of death before attainment of age x+1 in the middle of 2001. The 'base' year for these tables is thus 2000. Similar comments apply to the 92 series (mid-1992 to mid-1993) and the 80 series (mid-1980 to mid-1981).

Why are the exposed to risks for joint life cases so different for males and females in the smoker investigations?

One life may be a smoker and the other a non-smoker. This means that the smoker and non-smoker exposed to risks must be aggregated before the male and female exposed to risks are compared. (Ref: CMI Report 19, 105)

I am unable to reproduce the CMI results. When I multiply the exposed to risk at age x by the relevant mortality rate qx the expected number of deaths I calculate differs from the CMI’s figure. Why is this?

The CMI analyses use an initial exposed to risk and an age definition of age nearest. This means that lives are exposed from age x–½ to x+½. Since the mortality rate q applies as at the start of that year of age, qx–½ is used.

Can you provide an analysis of the temporary assurance investigation by ‘amounts’ as well as ‘lives’?

No. ‘Amounts’ data is currently only collected for the immediate annuitant and life office pensioner investigations. For all other investigations, including temporary and permanent assurances, we only have information on the number of policies. However, work is currently being undertaken to switch to a per-policy submission method for the mortality investigation, and this will enable us to carry out more detailed analyses. Find out more on the Per-policy data submission FAQs page.

Can you provide information on the correlation between the mortality experience and size of policy for the annuitant and pensioner investigations?

No. All data we receive is in an aggregated census format - we do not have details of individual policies. However, work is currently being undertaken to switch to a per-policy submission method for the mortality investigation, and this will enable us to carry out more detailed analyses.  Find out more on the Per-policy data submission FAQs page.

Can you provide any information on mortality experience broken down by occupation / socio-economic group / geographical location / cause / other?

No. We have no information on the make-up of the lives comprising the data, other than that they are all holders of policies with UK insurance companies. An investigation into mortality by cause of death used to be carried out, but this was discontinued in 1995. More information will become available once per-policy data collection is in place.

Can you suggest other possible sources of UK mortality information?

How should stakeholder pension plans be treated for data submission purposes?

Personal stakeholder pensions should be included in the same data return as for personal pensions data ie, investigation 22 for polices in deferment and investigation 24 for policies in payment. Currently there is no investigation for group stakeholder arrangements.

How will the reporting of annuities in payment be affected by pensions sharing on divorce?

Where the annuity record has been split so that both annuities are being recorded then the member’s pension can be recorded in the normal way (though at a reduced amount), and the ex-spouse’s policy should also be recorded as another active member (investigation 31 or 33). The point is that the spouse investigation (investigation 34) is recording the mortality of spouses after the death of their partner and the divorced spouse record should not be included in that investigation.

In the event of an office’s restructuring, for example through mergers, demutualisations or other transfers of business, should separate returns still be made for the original office numbers or should these be combined?

Provided the CMI is notified as to any changes in grouping, the way that policies are grouped is really a matter for the companies involved. The CMI aggregates the data prior to wider publication. Our main concern is to make sure that, overall, we collect as much accurate data as possible. At the company level, the main users of the data split are the offices themselves. This is reflected in the office specific reports produced for each office, and so if an office wishes to track the experience of each portfolio separately then it should continue to return data in the original way using the original office numbers.

What table is recommended for [the particular experience I am looking at]?

It is down to each individual actuary to choose the mortality table, and any adjustments to it, for the particular experience in question. We cannot “recommend” a table to be used.

What factors drive future movements in mortality improvements?

Possible factors include medical advances, lifestyle changes, etc. but unfortunately the CMI does not have information available.

Contact Details

If you have any questions about the CMI please email