The forward rates answer a question about arbitrage. Riding the curve asks the trading question sitting next to it. Suppose the curve slopes upward and you believe it will still look the same in a year. Why hold a one-year bill at 6.50% when you can buy a five-year at 7.00% and collect the higher coupon? A year later you sell as a four-year, and the same curve prices a four-year at a lower yield, which is a higher price than a four-year yield of 7.00% would give.
The strategy has three moving parts and the arithmetic keeps them separate. Carry is the coupon you collect over the holding period, plus what the coupons earn once reinvested. Roll-down is the price gain from the bond sliding to a shorter, lower-yielding point on an unchanged curve. Curve move is everything the market does that you did not assume. Carry and roll-down are known on day one; the third is the risk you are actually taking.
The critical thing to be clear about is the benchmark. Riding the curve is not compared against zero, it is compared against the one-year rate you could have locked in instead, and the honest comparison uses the same compounding basis on both sides. On the module's own bootstrapped India curve the one-year spot is 6.5016% quoted semi-annually, which is 6.6073% as an effective annual rate, and that 6.6073% is the number the ride has to beat.
Advertisement
How does the ride perform in two scenarios?
Two scenarios bracket the answer, and the gap between them is the whole theory of the trade.
If the curve is unchanged a year later, the ride works. Worked Example 6 computes the realised one-year return at 7.4574%, a pickup of 85 basis points over the one-year alternative. The pickup comes from the four-year point of the curve being lower than the five-year point that the bond was bought at, so the bond is repriced at a yield it never earned.
If instead the curve evolves so that today's forwards are realised, the ride earns exactly 6.6073%, the one-year rate, to the basis point. That is not a coincidence and it is the single most important sentence in this section. Forward rates are, by construction, the rates that make every holding-period strategy break even, so a curve that merely rolls up into its own forwards hands the rider nothing. Riding the curve is a bet that the curve will not do what the forwards imply, and an upward-sloping curve is not by itself evidence for that bet.
When does riding the yield curve lose?
Three conditions make the trade lose, and each has a number attached.
The first is the case above: the curve is already forward-implied, and the ride returns the one-year rate for four years of price risk. The second is an adverse move. Solve for the parallel rise over the year that pulls the realised return back to 6.6073%. On this curve the answer is 24.90 basis points, so a rise of more than about a quarter point wipes out the entire 85 bp pickup. A 50 bp rise turns it into a realised 5.7591%, some 85 bp below the bill. The asymmetry is the point. You are collecting 85 bp of pickup while carrying a bond whose modified duration at the sale date is 3.4395 years, and 3.4395 years of duration gives back 85 bp on a move of 24.7 bp, which is almost exactly the exact breakeven computed above. The third is transaction costs, which come straight off the top: a 10 bp round trip on a G-Sec leaves 75 of the 85, and on a less liquid state development loan the bid-offer alone can take a third of the carry before the curve has moved at all.
What else should you check before sizing the trade?
Two further points before you size anything. The pickup scales with the slope between the buy and sell points rather than with the level of the curve, so the same trade on a flat curve at 9% earns nothing while a steep curve at 4% pays well. And the breakeven above is a parallel move; a curve that steepens at the long end while the short end is anchored can take more than the parallel breakeven suggests, because the four-year point you are selling at moves further than the average. FI1.07 takes this to portfolio scale, where the ride becomes a positioning decision against a whole curve rather than one bond.