As the coupon goes down the duration (Macaulay, modified, effective) goes _ up _, not down. Scaling by 0,75 would be exactly the wrong thing to do. The duration might be 0,6 years instead of 0,5 years.
The inverse floater will have a longer duration than a fixed-rate bond, though exactly how much longer is difficult to say without doing a lot of work (e.g., Monte Carlo simulation).
Dividing it into three bonds as you suggest will definitely not work.
So, that 0,6 is a guess - does it mean that there is no way to make use of the fact, that the duration of a floater with e.g. C = 6mLIBOR is 0,5 year and we have some parameter k that scales it ( C = k x 6mLIBOR) ?
We have to compute Macaulay Duration traditionally - by hand ?
For that inverse floater - it confuses me much, because such portoflio consisting of above-mentioned bonds A, B and C is exactly the replication of that inverse. I see that i made a mistake, because portfolio’s duration should be weighted duration:
By the way, referring back to post #2, above, it’s an interesting exercise to determine why, exactly, the inverse floater will have a longer duration than the floater; indeed, it will have a longer duration than a fixed-rate bond. (In fact, it could have a duration that is longer than its maturity.)