[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Individual income protection experience 199598

LENGTHY DELAYS IN PUBLISHING results had become an
undesirable feature of the CMI income protection
(formerly PHI) investigation. It is therefore
pleasing to report that the income protection
(IP) investigation is quickly getting up to date on data
collection and analysis. This is evidenced by the results
of the individual IP investigation
for 199598, which
were sent to CMI members
in September 2000. This
article gives a short commentary
on the results. A
more detailed commentary
will appear in a forthcoming
CMI report.
The data
It is also pleasing to report
that the volume of data
submitted to the investigation
has increased substantially
over the course of the quadrennium. The
volume of in-force policy data increased by a factor of
44% over the period. Furthermore, the increases were
most marked in areas where data had been previously
sparse, eg deferred period 52 weeks and occupational
class 24. As this increase only happened in the later
years of the quadrennium, the effect is somewhat
watered down for this quadrennium but should have
a substantial impact in future if levels are maintained.
Changing office
The good news on data
volumes does have a flip
side. The experience of
offices varies enormously.
Figure 1 above shows an
example of this for inceptions
in respect of deferred
period, 13 weeks, class 1
males. Each of the seven
offices shown has at least
30 actual inceptions, so
the variation is not a result
of tiny data volumes. The
value of actual/expected
inceptions varies from
about 40% to 150%.
Clearly, if the proportion
of data contributed by
each office changes from
one period to the next,
any change in underlying
morbidity may be masked as a result. The most dramatic
changes to the mix of business are caused by
offices which join or leave the investigation from one
year to the next. This article therefore shows the ‘raw’
results for all offices, split by occupational class, for
199598 but also shows a comparison of results for the
last three available quadrennia for a subset of those
offices that contributed to the investigation throughout
the 12-year period. This eliminates a major cause
of distortion but other factors may still distort any
underlying trends in morbidity, eg changes in data volumes
per office, changes to underwriting and claims
control practices, etc.
199598 results by occupational class
The methods of analysis are those described in detail
in CMIR15. These compare actual claim inceptions
and terminations, by recovery and death, with those
expected on the basis of the graduations of the male
Standard experience for individual IP policies for
197578 as set out in CMIR12. This comparison basis
is referred to as SM197578.
CMI has been collecting data split by occupational
class since the start of the 199194 quadrennium.
Offices submit data using their own internal class coding.
These are then converted by CMI to the most
appropriate of the four CMI standard classes for analysis
purposes. The classes are described fully on p4 of
CMIR17 and range from the non-manual class 1 to
the heavy manual class 4.
The data analysed by occupational class is a subset
of the overall (aggregate) data consisting of UK policies
subject to no identifiable underwriting exclusions
and no special benefit types (eg lump sums). This is
known as the Standard* data.
The Standard* data is subdivided between data from
the four occupational classes and a diminishing ‘class
unknown’ category where offices cannot supply occupational
class information.
Prior to 1991 offices submitted no occupational class
information but were asked to indicate whether a policy
was occupationally rated. This indication field has
been retained in the data requirements alongside the
occupational class field to see how the two correspond.
The old occupational rating field was used to define the
Standard data subset along with the same criteria used
to define the Standard* subset. The Standard subset has
been used for many years to reduce heterogeneity in the
aggregate data. It is itself therefore a subset of the newer
Standard* subset, but leaves out those policies which are
shown as occupationally rated. Comparison of the
occupational rating field with the occupational class
field for recent data shows that not all Standard data is
class 1, but the great majority appears to be class 1. The main use of Standard data is for historical comparisons
with results prior to 1991, where occupational class
information was not collected.
Table 1 (across) shows the overall inception results
for the 199598 Standard* experience for both males
and females, split by occupational class. Table 2 above
shows the equivalent results for termination by recovery.
Terminations by death are far fewer in number
and are not discussed here, for reasons of brevity.
The key features of the results are as follows:
°ª female inception experience continues to be much
heavier than male experience;
°ª inception rates generally increase with occupational
class from class 1 to class 4;
°ª there appears no evidence of any pattern for recovery
rates to vary by occupational class;
°ª female and male recovery rates are much more
comparable than inception rates.
Comparison with earlier experience
The above information relates to 199598 all office
data. It is of natural interest to actuaries how this compares
with previous experience. As explained above,
the change in the mix of contributing offices can lead
to misleading comparisons. For this reason we have
chosen to isolate the combined experience of those
offices that were present throughout the period
198798 and compare the experience of three quadrennia.
The Standard subset is used which enables us
to include data prior to 1991 and eliminates distortions
caused by offices starting to supply occupational
data at different times.
Table 3 right shows, for each combination of sex and
deferred period, the number of actual inceptions and
the percentage of actual/expected (A/E) inceptions for
the group of offices in respect of each quadrennium. It
also expresses the A/E figures for the two later quadrennia
as a percentage of the 198790 figure. The
results for males show a decreasing trend for the shorter
deferred periods and an increasing one for the longer
deferred periods. For the 13-week deferred period,
inception rates reduced in 199598 after increasing in
199194. For females, decreasing trends are evident for
the four-week and 13-week deferred periods but the
picture is less clear for the other deferred periods.
Table 4 below shows the equivalent figures for
claim termination by recovery. This indicates a trend
towards declining male recovery rates for all deferred
periods apart from 1-week. The female results are less
A final word
The investigation would not be possible without the
commitment, resources, and hard work put in by the
offices that contribute data. We are very grateful to
the offices concerned. Clearly, armed with up-to-date
data, there remains great scope for further research
into this complex and interesting area of the insurance