A gilt fund runs duration 6 at a 7.2% yield. The manager expects 75 basis points of rate cuts and wants to go to duration 9. Build the one-year return table for cuts, no change and a 50 basis point rise, and decide.

Extending to duration 9 adds about 2.25 points of return if 75 basis points of cuts arrive and costs about 1.5 points if rates rise 50; with no change it makes no difference. At duration 6 the fund earns 11.70%, 7.20% or 4.20%; at duration 9, 13.95%, 7.20% or 2.70%. The decision turns on whether the cuts are already in today's 7.2% yield. Extend part of the way, to about 7.5, unless the manager can show the market has not priced them.

Think of a fixed deposit you could sell to someone else. You earn the interest while you hold it, and if new deposits start paying less, yours becomes worth more to a buyer. A bond fund's one-year return is roughly its yield plus a price change of minus duration times the change in yield. Modified durationThe approximate percentage change in a bond price for a one percentage point change in its yield, with the sign reversed. is the lever: at duration 6, a one point fall in yields lifts prices by about 6%.

So the table is simple arithmetic. Duration 6 with 75 basis points of cuts: 7.2 plus 6 x 0.75, which is 11.70%. Duration 9: 7.2 plus 9 x 0.75, 13.95%. With a 50 basis point rise, duration 6 earns 7.2 minus 3.0, 4.20%, and duration 9 earns 7.2 minus 4.5, 2.70%. Extending by 3 years of duration adds 3 points of price move for every 1 point change in yields, in both directions.

Because bond prices move on surprises, not on events. If the market already expects 75 basis points of cuts, today's 7.2% yield has those cuts built in, and the cut arriving moves prices very little. The payoff table above assumes yields fall by the full 75 points from here, which only happens if the cuts are not yet priced. The first question to put to the manager is what the forward curve and the overnight index swap market already imply, and how his view differs from that.

Then weigh the asymmetry. Extension gains 2.25 points in the cut case and loses 1.5 in the hike case. Ignoring the no-change case, it pays off on average if the cut case is more than 40% likely against the hike case. That sounds easy to clear, but the gain only counts if the cuts are a surprise, so the real hurdle is the probability that the market is wrong, which is far lower than the probability that cuts happen.

Take a measured position rather than the full bet. Moving to about duration 7.5 captures half the extra gain, about 1.1 points if the manager is right, and halves the extra loss to about 0.75 points if he is wrong. Keep room to add if data confirm the view. Gilt fund investors did sign up for rate risk, so a duration call is legitimate; what they did not sign up for is a single forecast deciding a third of their year's return.

State the limits of the table. Convexity adds a little in both directions and slightly favours the longer portfolio, roll-down earns extra if the curve is steep, and a curve that twists rather than shifts will not match a single duration number. For a desk interview the linear table is the right first answer, followed by those caveats in one sentence each.