[Quantlib-dev] proposal for inflation seasonality

[BL BL <blq...@gm...>]

Hi,

while in need of the seasonality described in the Lehmann Paper "Inflation
Derivatives explained" as can be found here
http://the.earth.li/~jon/junk/Kerkhoff.pdf i specialized the
multiplicativepriceseasonality class.

When overwriting the seasonal correcttion and the seasonalityFactor Methods
i found that the seasonal factor obtained from seasonalityFactor() was not
reproduced when i calculated the ratio of the undajusted forecasted fixing
(obtained by calling setSeasonality("myseasonality") devided by the
unadjusted value (obtained with setSeasonality() thus removing seasonality).

But this is a basic requirement. I was able to track this down to a
(seemingly) mismatch in the call to the index forecast method. Which is in
intlationindex.cpp at line 186 (QL1.5). And the call to the
seasonalCorrction() at inflationtermstructure.cpp at line 173.

So seasonal Correction is always called with d-uselag. While the base for
the forecasting might differ from that.

In my view it should never differ. Adjusting the zero-rate rate should
always reproduce the factor. And so the time implicitly used in this zero
rate adjustment should always be the time used in the application of the
zero rate for forecasting.

To make a long story short i would suggest either coying the behavior in
the forecast so to always pass the same baseDate to the seasonality as the
one used in forecasting.

Or, if we keep the call to seasoaalityCorretion() as it is, that is it gets
d-uselag, it seems to me a better default would be the first day of the
inflation reference month. That would at least fit well in case of an
uninterpolated HICP Index and an application of the kerhoff paper
seasonality.

Any Suggestions?