The option a farm turns down has had a box of its own on the partial budget form for decades. University of Maryland Cooperative Extension, in a fact sheet issued in 1991, lays the form out in four parts: “The partial budget has four categorical parts: additional income, reduced costs, reduced income and additional costs”. The third of those, reduced income, is the money that stops arriving because of “enterprises being eliminated, reduced in size or output”. Naming what the farm gave up is not an advanced technique. It is a box on a form, and what this page asks for is that box, filled in and kept.
The big decisions get settled with a figure in somebody’s head and none on paper. Somebody weighed renting the ground out against cropping it, or finishing the animals against selling them at weaning, or hiring the contractor against buying the machine, reached a conclusion, and moved on. Prices turn a few months later, somebody asks whether it still holds, and the argument restarts from nothing, because the number that decided it was never written anywhere. Where it came from was never written either, so there is no way to tell whether it has aged.
What the sheet asks for is not the calculation the farm should have run. It is the figure it actually used, the place that figure came from, and the day it was used. Three rows, six columns, one page. Decisions taken on habit, with no number behind them at all, get a row too, and those turn out to be the rows worth having.
Why no statement the farm receives will ever show this figure
Nothing was paid, so nothing recorded it. Conab, the Brazilian agency whose cost of production series covers the country’s main crops, files this quantity among the implicit costs, the ones for which no actual disbursement occurs, alongside depreciation, and defines it as the value a given factor could receive in some alternative use.
The United States Department of Agriculture arrives at the same place from the other side in its own costs of production handbook, which states that an average operator could expect the machinery, equipment, breeding livestock and labor the farm already owns to return each year what those same inputs would have earned in alternate uses. That expected return is real enough to be charged against the crop, and it is also invisible, because nobody ever wrote a check for it.
Two agencies, two hemispheres, one consequence at the gate. The bank statement will not hold this figure, the invoice folder will not hold it, and no accounting package will produce it, because there was never a payment to record. That is why the sheet belongs in the finance axis rather than in a drawer of plans: it is a money record with no transaction behind it, and a page written on purpose is the only place it can exist.
What the form has asked for since long before any of it was software
Name the alternative first, then date the page. Penn State Extension opens its partial budget procedure with exactly that instruction, and gives the reason: “It is important to have a clear understanding of exactly what alternative is being considered, especially when sharing your results with other decision-makers or discussing your options with your banker or other adviser”. The second of the four questions the same guide says every partial budget has to answer is “How much current income will be lost or reduced?”, which is the turned-down option in question form.
Dating the page comes in the same step, as “Add a date to your analysis”, and the guide attaches the clause this whole page runs on: keep the evaluation “even though you may decide not to make any changes now”. The Maryland worksheet had already built that in, printing a line for a name and a line for a date across the top of the blank partial budget form, above the line where the proposed change gets stated. Two public extension services, nearly thirty years apart, both put the same two things at the top of the page, and neither one has anything to do with arithmetic.
What the six columns hold
One row per decision, and nothing on the row that needs looking up. Two of the six are already printed on the blank worksheet the Maryland fact sheet hands out, the date and the line where the proposed change gets stated. The other four are what turns a one-off analysis into a record.
| Column | What goes in it | Where it comes from |
|---|---|---|
| The decision | One line, in the words used at the time | The person who settled it |
| The date | The day it was settled, not the day the sheet was written | The contract or invoice that followed, when memory is unsure |
| The option turned down | Named in full, never “the alternative” | The same conversation |
| The number that ruled it out | The figure with its unit, and beside it how far it sat from the level that would have reversed the call, or the words “no number” | Whatever was on the table that day |
| Where the number came from | Contract, written quote, published tariff, advertisement, or a person by name | The document itself, filed or referenced |
| When it comes back to the table | A date, on the calendar the farm already keeps | The rows that stand on a price |
Why the sheet records the figure instead of computing the right one
Computing it properly is a piece of research, and the bodies that do it say so in writing. The FAO handbook that national statistical offices work from defines the quantity as the return the capital would have earned in the next best alternative, then states the limit in the next breath: “This estimate can be difficult to compute and subject to error”. It “involves judgment on the part of the statistician on several dimensions”, and the handbook names two of them, deciding the market price of the asset when few of them trade, and deciding what rate of return to assume.
A refereed paper shows the size of the job on a single farm decision. Three researchers priced the choice between buying pest control products and multiplying them on the farm, for a soybean case in Brazil, and published it in Revista de Economia e Sociologia Rural in 2025. The definition they work from is the one every methodology uses, “the net income that the factor generates in its best alternative use”, and then, to make it computable at all, they write the assumption the whole comparison rests on: “Assuming that productivity in both management approaches is similar, we consider only the costs for this analysis”.
That is a published estimate, reviewed by other economists, standing on an assumption its authors had to declare in order to get a figure at all. A number worked out at the kitchen table at nine in the evening stands on an assumption too. The only difference the sheet creates is that the second one gets written where somebody can read it later, which is the whole of what it is for.
The column that decides whether the row is worth anything
Where the figure came from, in writing, next to the figure. Penn State sorts numbers by how far they can be trusted before the comparison is even done: “Hard numbers are those items that have cost or income values you can assign with a high degree of certainty”, against soft ones the manager is less sure of. For the soft ones the guide says to “adjust them to see how much the estimate would need to change before you would change your mind about adopting the alternative”, which is only possible if somebody can find them again.
On the sheet that becomes one column, and the entry is whatever is honestly true: a signed contract, a written quote, a published tariff, a printed advertisement, or a person at the gate. That last one belongs on the list, not below it. The same guide tells the reader to “Talk with producers who have experience with the alternative you are considering”, so a figure that came out of a conversation is a legitimate figure. A figure whose origin nobody wrote down is not, because when the price moves there is nothing to go back to.
Some of these figures come from inside the gate rather than from any document outside it. A decision to shrink one activity is usually settled by a cost per unit that came out of an allocation, and when that is what happened, the source column says so by name. That sends whoever reads the row a year later to a declared split rule, where the question is whether both sides of the comparison were measured with the same ruler, which is a different question from whether the arithmetic was right.
Why the figure can only be captured on the day
After the money is committed, the official methods stop calculating it, and they are right to. Argentina’s Ministerio de Agricultura, Ganadería y Pesca states the rule flatly in the manual behind its published cost and margin tables. Money is an input like any other, and its cost is not being able to use it in activity B because it is tied up in activity A, a cost the manual measures through the interest on that capital.
The manual then splits the calculation by direction. Looking forward, before the activity happens, the cost of the money about to go in has to be counted. Looking back at a result already obtained, the chance to invest in another business no longer exists, because the money has already gone into the business under analysis, and in that case the manual says that interest is not to be computed at all.
Read that toward the sheet and it is the argument for writing on the day. The figure that ruled the alternative out is a live quantity for exactly as long as the decision is open, and once the money moves, the methodology that a national ministry publishes its own margin tables under stops counting the equivalent quantity in its own accounts. Nobody is being careless.
There simply is no later procedure that recovers what the farm was looking at that morning, and the figure comes back from memory or it does not come back at all. That is why the sheet asks for the last three decisions and not the last thirty: three is what memory still reaches. Every decision settled from here on goes onto the sheet the day it is settled, while the figure is still live.
The other direction has its own page. A written investment plan holds the condition that has to come true before an item leaves the paper, written before the money is spent. This sheet holds the figure that already turned something down, written after. Both are one page each, and a farm that keeps only the first can say what it intends to buy and not what it has already refused.
How far the figure sat from flipping
Write the distance, not only the figure. The Maryland fact sheet runs that arithmetic backwards in its own teaching case, and it is the most useful line in the document. A manager there compares two crops on the same ground and comes out ahead by a margin that reads as comfortable.
The fact sheet then divides that margin by the assumed price, which turns it into the fall in yield that “would cause you to be indifferent to making the change in the operation”, and divides it by the assumed yield, which turns it into the fall in price that does the same thing on its own. Two divisions, and the comfortable margin has a distance printed beside it. That distance goes in the same column as the figure, beside it, because the two are read together or not at all.
The two divisions are what transfers. A decision that survived by a wide margin and a decision that survived by a hair are indistinguishable in memory a year later, and on paper they are not. Writing how far the number sat from the level that would have reversed the call is what tells the farm which rows to watch and which to leave alone.
The rows that stand on a price are the ones that need a date. That date goes on the calendar the farm already keeps rather than on a second one, and it works when a name sits beside it, which is the finding that a scheduled review date is built on. The lowest price the farm can accept turns up on these rows more often than any other figure, and it moves with the cost lines behind it, which is why the break-even price carries a date of its own.
The rows with no number at all
They are the finding, not the gap in it. Farm records hold what was done. A 2024 review in the journal Precision Agriculture surveyed on-farm recordkeeping across paper notebooks, spreadsheets, database forms, web tools and phone apps, and describes the oldest of them, the farm logbook, as “the archive of field work by date, time, and person who conducted the operation”.
The review credits the logbook with the thing this page is after, that these books “aid in planning, decision-making, and troubleshooting by offering historical data for evaluation and analysis”. Read the two sentences together and the shape of the hole appears. What gets archived is the operation, by date, by hour, by person, and what gets left out is the choice that put the operation there. Not one category in that survey carries the option that was rejected, because none of them was built to.
A farm that reaches a row and finds no figure has not failed anything. It has located a decision it takes on habit, and habit is a decision rule like any other: it is dated, it came from somewhere, and it can be wrong for years without anybody noticing, because nothing about it is written where a second person could check it. Marking those rows is the fourth movement of farm management doing its job, which is to send what the first three produced back into the next decision.
Where to start
One to two hours, alone, with nothing on the table but the last twelve months of memory and whatever the farm already filed. One page, three rows, six columns, dated at the bottom.
Fill this in honestly and at least one row will show that the figure came from the party selling the thing the farm bought. That is neither a scandal nor unusual: a dealer’s quote is often the only number available on the day, and a decision taken on it can be perfectly sound. What the column does is make the fact legible a year later, when the same dealer calls about the next machine and somebody can see, in writing, that last time the farm let the other side of the table supply the figure that closed the argument.