In at least one paper I've seen on this subject, month was fixed as a 30 day period. But even without that, as an approximation, months work pretty well for reporting the rough increase. The difference between 29 and 31 days will get swamped in the error margin for anything less than six months or so. And since most months are 30 or 31, and many alternate, in practice, this will be very close to 30.5 days, so the difference will be negligible.