A duration-neutral position is constructed so that the total portfolio DV01 is zero. Gains from one leg offset losses from the other when rates move in parallel. This isolates the trade's profit-and-loss to changes in the yield curve's shape rather than its level.
The construction process:
Example: a 2s/10s flattener with the 10-year having a DV01 of 820/million and the 2-year having a DV01 of 195/million requires a hedge ratio of 820/195 = 4.2x. For every 1 million of 10-year notes sold, approximately 4.2 million of 2-year notes must be bought.
Duration-neutral construction is essential because parallel shifts in the yield curve are far larger (in terms of P&L impact) than slope or curvature changes. Without duration-neutrality, a curve trade would be dominated by directional rate moves, swamping the intended spread exposure.
The Salomon Brothers yield curve primer emphasizes that virtually all professional yield curve trading is done on a duration-neutral basis. This applies to curve trades, barbell and bullet comparisons, and butterfly spreads.
The most common error in curve trading is to match the size of the two legs instead of their risk.
Take the flattener above and build it with equal notional. Sell 10 million of the 10 Yr and buy 10 million of the 2 Yr. The short leg carries 8,200 of DV01. The long leg carries 1,950. The position nets to about 6,250 of DV01 short.
That is not a curve trade. It is a short duration position with a small curve trade attached. If yields fall 20 bps across the whole curve and the spread does not move at all, the position loses roughly 125,000 dollars on a view that was never taken.
The numbers get worse as the legs move further apart. A 2s30s trade built on equal notional is almost entirely a directional bet, because the 30 Yr leg can carry five times the risk of the 2 Yr leg. The wider the maturity gap, the more the notional and the risk diverge.
Matching DV01 leaves the two legs at very different sizes, and that difference has to be funded.
The flattener above sells 1 million of 10 Yr notes and buys 4.2 million of 2 Yr notes. Roughly 3.2 million of cash is required beyond what the short leg raises. On a rates desk that gap is closed in the repo market, and the cost of closing it is part of the trade.
This creates a genuine choice.
Most rates trading uses the first. Portfolio managers working within a cash constrained mandate sometimes use the second and then hedge the leftover duration separately.
A duration-neutral position is not a hedged position. Three exposures survive.
The curve view itself. This is intentional. The trade is built to profit from the spread moving, and it will lose if the spread moves the wrong way. Removing directional risk concentrates risk rather than reducing it.
Convexity. The longer leg is more convex than the shorter one. A flattener that is long the 2 Yr and short the 10 Yr is therefore short convexity, and a large parallel move in either direction costs money even if the spread does not budge. Over a few basis points this is negligible. Over a sharp repricing it is not.
Drift. DV01 changes as prices change and as the bonds age. A trade that was neutral when it was struck will not stay neutral. The hedge ratio has to be recalculated and the position rebalanced, and each rebalance costs a spread.
It depends on the size of the yield move and the tolerance of the desk. A common approach is to set a threshold, such as allowing net DV01 to drift to some fraction of the original leg risk before adjusting. Rebalancing too often pays away the trade's edge in transaction costs. Rebalancing too rarely lets a directional exposure build up unnoticed.
Yes, and this is the more common case for asset managers. The target is not zero DV01 but the DV01 of the benchmark index. The manager is then neutral to parallel moves relative to the benchmark while taking active positions along the curve.
Yes. The same arithmetic applies to futures and swaps, with one extra step for futures. A futures contract derives its risk from the bond that is cheapest to deliver into it, so its DV01 is the risk of that bond adjusted by the contract's conversion factor.