Re: [lmi] questionable code checking for e_currbasis in ihs_acctval.cpp

[Greg Chicares]

On 2008-03-24 17:10Z, Greg Chicares wrote:
> On 2008-03-24 04:31Z, Vadim Zeitlin wrote:
>> 
>> But maybe the real intention was
>> 
>>      e_currbasis != ExpAndGABasis

Fixed all five occurrences in 'ihs_acctval.cpp' 20080325T1259Z.

They were all in identical early-exit conditions in experience-
rating functions:

        ...
        ||  !Input_->UseExperienceRating
        ||  !e_currbasis == ExpAndGABasis
        ...

which aren't called (or whose results aren't used) except on the
current-expense and current-interest basis, as is documented,
more or less, here:

'group_values.cpp'
            // Calculate next year's k factor. Do this only for
            // current-expense bases, not as a speed optimization,
            // but rather because experience rating on other bases
            // is undefined.

though the actual condition is stricter:

            if
                (   cells[0].UseExperienceRating
                &&  e_currbasis == expense_and_general_account_basis
                &&  0.0 != eoy_inforce_lives
                )

> We'll have to test this correction carefully, even though it's
> "obviously" right. I think we're okay because experience rating
> would be shown in practice only on current-basis supplemental
> illustrations, but we will have to revise regression tests.

Not just "in practice": experience rating indeed isn't performed
on any other basis. Therefore, regression tests are unaffected.

It wouldn't be unreasonable to relax the condition someday and
permit experience rating on a zero-interest basis, as the comment
above may be read to permit, and which would have unmasked this
defect had we not corrected it.