DISCUSSION 581
could be argued that all extra persister lapses will occur immediately when
reversion is denied and that no additional lapses will occur after that point.
This would confine the blended mortality problem to one cohort of reverters
but would leave the problem of choosing the blended mortality level. The
problem expands if one assumes that there will be additional persister lapses
in all durations after the first reversion opportunity. Assuming higher persister
lapses at the first reversion opportunity and lower persister lapses afterward
complicates matters further--those lower lapses could be considered as negative
reverters, perhaps in a high-risk class.
In any case, it should be clear that assuming different lapse rates for per-
sisters and reverters poses some serious challenges for the pricing actuary.
2. Mr. Bakos's approach to the above-described complexities is to assume
that mortality levels and persistency levels operate independently of each
other. If we read his comments correctly, he believes that one can set the
persister premium at a level high enough to eliminate deficiency reserves and
not worry about the effect of persistency on the viability of those premiums.
It seems clear to us that setting persister premiums at a table 4 level as he
suggests would expose the company to the same cycle of lapses by the better
risks (table 3 or better), leading to higher sustained mortality, which, in turn,
would lead to losses or higher persister premiums.
3. A proposed alternative to these complicated formulations is the use of a
"conservation-of-total-lapses" principle. Although we had difficulty following
Messrs. Gould and Porter's calculation, it appears that this approach solves
for the persister-class lapse rate by establishing a lapse rate for the reverter
class and assuming that the mortality rate for the two classes is the same (much
as we solved for the persister-class mortality rate by assuming a reverter-class
mortality rate and equal lapse rates for the two classes). This assumption does
not seem appreciably better than our assumption that the lapse rates are the
same. Having to assume that mortality rates are the same for both classes in
order to arrive at this assumption is one flaw. Also, the conservation-of-deaths
principle works because people do not choose to die; so as long as you insure
the same class of risk, total deaths should be the same. The conservation-of-
lapses principle does not work because people can choose to lapse depending on
the premium scale they are paying; thus, total lapses would not necessarily be
the same.
4. In any case, Messrs. Gould and Porter's discussion shows that there ap-
pears to be no substantial difference between persister mortality calculated
using our admittedly convenient lapse assumptions and that calculated using
their approach with separate lapse rates for persisters and reverters. The maxi-
mum differential is roughly 7 deaths per 100,000 and is often much less than
that. This differential seems especially small in light of the approximate nature
of the other assumptions that must be made in pricing this product. These find-
ings corroborate our conclusion that relative lapse rate differentials have only
a minor effect on persister mortality.