Dedication buys a portfolio whose cash inflows arrive on the same dates and in at least the same amounts as the liabilities they fund. Nothing is forecast, nothing is rebalanced, and interest-rate risk disappears because no asset ever has to be sold or reinvested at an unknown rate.
How is cash flow matching done, step by step?
The construction runs backward from the last liability, and it has to run backward for a reason that is easy to miss. A bond bought to fund the final payment also pays coupons on every earlier date, so it contributes to every earlier liability. Solve forward and you would double-count those coupons. Solve backward and each step knows exactly what the later bonds have already contributed.
The recipe, on a schedule of liabilities at times 1 through n:
- Buy enough of the bond maturing at time n so that its final cash flow, principal plus coupon, exactly meets liability n. That fixes its face value.
- Move to time n-1. Subtract the coupons the bonds already bought will pay at that date. Buy enough of the bond maturing at n-1 to meet the remainder.
- Repeat down to time 1, at each step subtracting every coupon already committed.
Advertisement
What goes wrong with cash flow matching in practice?
Two things break the clean version. The first is that the universe rarely holds a bond maturing on every liability date, so surplus cash has to be carried forward from an earlier date and reinvested at an unknown rate. Prudence requires assuming a conservative reinvestment rate, and every basis point of conservatism is paid for at the front. In Worked Example 1 the missing four-year maturity is the entire reason dedication costs more than immunisation: assuming 5.00% on the carried surplus rather than the curve's implied forward of 7.400561% requires ₹0.268685 crore of extra cash at year three, which is ₹0.219327 crore in present value, or 34.35 basis points of the liability. Every rupee of that gap is traceable to one assumption.
The second is that a real dedication problem is an optimisation, not a recursion, once the universe is large and lot sizes are integers. The backward recursion gives the exact answer for one bond per date; a linear program minimising cost subject to a non-negative cash balance on every date gives the answer when there are forty candidates and minimum lots. Both give the same object, and the recursion is what you must be able to do by hand, because it is the only version whose answer you can check.
One practical warning about the optimiser. A cost-minimising solver will happily fund a schedule with a portfolio that leaves a zero cash balance on every date, and a zero balance is a portfolio with no room for a single missed coupon or a single day's settlement delay. Real dedication mandates therefore carry a minimum cash buffer on every date as a constraint, and the buffer is a policy number the trustee sets rather than an output the solver produces.
What are the strengths and costs of dedication?
Dedication's virtues are unusual and worth stating plainly. It needs no rebalancing, it needs no view, it survives any curve shape and any twist, and its risk report has one line on it. Its costs are equally plain: it demands a bond maturing near every payment date, it forgoes any return above the matched yield, it is expensive to unwind, and it locks the portfolio out of a rally.