Marginal Return on an Input

Every input decision on a farm is a decision at the margin. Nobody is choosing between fertiliser and no fertiliser; they are choosing whether the last thirty kilograms of nitrogen, the second fungicide, the extra tillage pass or the higher seeding rate earned back what it cost. Divide the extra revenue by the extra cost and the answer is a ratio, conventionally read as dollars back per dollar spent.

The rule agronomists use is that an input should return at least two to one before it is recommended, and the second dollar is not timidity. This calculation is done on an AVERAGE response, drawn from trials across seasons and soils, while any particular field in any particular year lands somewhere on a wide distribution around that average — and a fair share of the time the response is zero, because the disease did not arrive or the nitrogen was already there. The margin above one-to-one is what pays for those years. An input sitting at 1.2:1 on the trial average is, in practice, a coin toss.

The word marginal has to be taken seriously, because input response curves flatten. The first increment of nitrogen on a deficient soil might return five dollars for one; the increment that takes the crop to its yield plateau might return sixty cents. That is why the yield-maximising rate and the profit-maximising rate are different numbers, and why the profitable rate falls when fertiliser rises or grain falls — the optimum is wherever the last unit's return crosses one-to-one, and both prices move that crossing every year. A grower applying last year's rate at this year's prices is answering a question nobody asked.

Finally, the increment has to be a real increment. Comparing a treated field to an untreated field next door measures the soil type, the seeding date, the drainage and the previous crop at least as much as it measures the input, and the difference is usually assigned entirely to the product — which is precisely what the product's marketing hopes you will do. Strip trials laid out across a field, replicated and preferably repeated, are the cheapest honest answer available, and a yield monitor makes them nearly free.