There is no correct way to divide a cost that serves two activities, and the institutions that publish cost methodology admit it in their own manuals. What each of them does instead is pick a basis, write down which basis it picked, and print that alongside the figures. A farm can copy the second half of that even when it cannot settle the first: choose what every shared line gets divided by, write it on a sheet with the date, and do the arithmetic afterwards.
The shed, the tractor and the accountant serve the cattle and the crop at the same time, and sooner or later somebody has to say which of the two is paying for itself. The shed gets divided by area, and the cattle come out cheap. The following cycle a different person divides it by revenue, and the cattle come out expensive.
Both figures are arithmetically sound, they answer different questions, and neither basis was written down before the answer appeared. The farm now holds numbers for both activities, believes them, and cannot set one cycle against the next, because the ruler changed in the middle and nobody recorded the change. When the decision to shrink an activity finally arrives, it gets taken with whatever the most recent basis produced.
The manual that national statistics offices use for this does not hide the difficulty. The Food and Agriculture Organization, in its handbook on cost of production statistics, calls the split a necessary step and then says of the methods available that “While there are many approaches for doing this, none are perfect and all of them can result in biased or faulty estimates.” That sentence is the reason this article exists. A problem with no correct answer still has to be answered every year, and the only thing that can be got right about it is the record of what was answered.
Do the bodies that publish cost methods agree on a basis?
No, and the disagreement is on the record, in the methodology each body publishes. For the overhead that cannot be attributed to any one product, the FAO handbook says it is generally accepted that such expenses be allocated “based on the relative contributions to the whole farm net margins”. The United States Department of Agriculture has used two different bases in the same account, general farm overhead “allocated by relative value of production” under its older method and the “Same farm costs as in USDA method allocated by relative gross margins” after 1995.
Conab, the Brazilian federal supply company, divides buildings and installations by a physical share instead, the occupancy rate of the asset, the percentage of use of that asset in a given crop, taken from the average use of the tractors in that crop. Three public bodies, three continents, three bases. What all three do the same way is state which one they used.
What changes when the basis changes?
The number, and sometimes which activity looks like the good one. The FAO handbook works a case from the Philippines that fits on one line: a rice grower with three parcels, two hectares of rice, one hectare of rice and half a hectare of mango, using one hand tractor across them. Divide the tractor by total farm area and the focus parcel takes 57 percent.
Ask first whether the tractor ever went to the mango, find that it did not, divide by rice area only, and the same parcel takes 67 percent. Same machine, same year, same grower, two defensible rules, and the ten points between the two figures are ten points of a tractor moving from one activity to the other. Nothing about the facts decided that. A question decided it, and the question was asked by whoever was holding the pen.
The same herd, two bases, two results
Cattle carry a version of the same problem, and it has been worked out on paper. Aranha, Dias and Itavo, writing in the Revista de Economia e Sociologia Rural in 2016, took a worked full cycle beef example and ran the indirect costs through two bases. Under the traditional one, the pile is divided by head, so a calf carries the same cost as a grown cow. Under a basis built on animal units, which convert animals of different ages and sexes onto one scale of live weight, the cost follows the weight.
The comparison the authors report is the useful part for a farm: on the head basis the young light animals came out overvalued and the older heavier ones undervalued, the inventory value differed between the two, and so did the gross result on the sale. This is one worked example from the accounting literature and not a field trial, so what carries across is the direction of the error and not any figure in it.
The basis that moves when the price moves
Anything divided by revenue moves with the market, which is the one property that kills a comparison between two cycles. Gazzarin and Lips, at the Swiss federal research institute Agroscope, reviewing the methods used in farm activity accounting in 2018, sort the available bases into two families. The physical ones are working hours, tractor hours, area and livestock units.
The monetary ones are book values and turnover per activity. On the monetary family their verdict is flat. Loading each activity with the share of the cost it can afford to carry is of little use for analysis, because the figures that come out of it do not show cleanly what happened inside the farm, the level of cost ending up set by the selling price.
The physical family gets no blanket approval from them either, and the reason is a different one: what they ask of a physical measure is that it be the one closest to whatever actually causes the cost, with the livestock unit and the hectare as their own examples, and the difficulty they name is finding one that is both close to the cost and cheap to measure. That difference is the one a farm needs. A badly chosen physical basis is wrong and stays wrong, which is something a farm can find and fix. A monetary basis moves on its own between one cycle and the next.
Read that against a farm with a shed over both enterprises. Split the shed by revenue and a good price year loads more of the shed onto the enterprise that sold well, so its cost per unit rises in the year it earned most, while nothing physical changed in the shed, in the hours or in the herd. Two cycles priced differently are then not comparable at all, and the comparison will still hand back a winner. This is also why a revenue basis and the break-even price sit badly together: the floor is supposed to be the fixed thing the price is tested against, and a revenue split lets the price move the floor.
The one rule that really is a rule
Match the basis to the nature of the expense. It is the single instruction the FAO handbook states as a requirement rather than an option, that “Allocation keys should be related to the nature of the expenses to be allocated”, and it comes with a negative form any farmer can check without a spreadsheet. The handbook tells the statistician they “should not use land area to allocate non-specific livestock expenses and animal counts to allocate non-specific crop expenses”.
Run that over your own sheet before anything else. Vaccine fridge electricity divided by hectares is that error. Fence repair on the crop side divided by head is that error. A shed used four months by one enterprise and eleven by the other, divided equally because there are two of them, is a third version of it, and the fix is not a better formula, it is picking a basis that has something to do with what the shed is doing.
What a split rule can be built on
From what the farm already counts, which is the reason the FAO handbook’s table of allocation methods is worth having on the table while the sheet is written. Every row of it names the assumption you take on when you choose that basis, and the assumption is the part that gets forgotten.
| Line to divide | What it can be divided by | What choosing that assumes |
|---|---|---|
| Fertilizer and crop protection | Crop-specific application rate, planted area, production quantity | Using area assumes the same rate across crops |
| Machinery, fuel and lubricants, electricity and heating | Hours or days of use with engineering data, planted area | Using area assumes the same intensity of use across products |
| Buildings | Harvested area, head of cattle, production quantity, production value | Same frequency of use across products, and harvested rather than planted area for a harvester |
| Labor and feed | Head of cattle, labor intensity by task, feed rates by cattle type | Same intensity of use across cattle types |
| General management, taxes, insurance, permits and licenses | Planted or harvested area, head of cattle, cattle value | Space occupied or value added, and only comparable within products of the same type |
Two of those rows depend on numbers this cluster works out elsewhere. Hours of use come from the repair bill for one machine, which is where the hour meter reading gets turned into a cost per hour. The annual slice of a machine is fixed cost by definition and its size comes from the useful life you declared, which is the other declared standard this sheet leans on.
The activity that looks like it pays for itself
An enterprise budget with no share of the common pile in it. Oklahoma Cooperative Extension Service names the pile and then names the practice: “Costs that are difficult to allocate to individual enterprises (telephone, taxes and accounting services, electricity, etc.), are called overhead costs and may be variable or fixed in nature.” and “Overhead costs are included in whole farm/ranch budgets, but are sometimes excluded from enterprise budgets.”
The consequence shows up in the line the same fact sheet uses to read a budget, that “When returns to overhead, risk and management are positive, the enterprise is profitable and self-supporting.” Leave the pile out of every enterprise and every enterprise reads as self-supporting while the farm as a whole does not, and no line in any of the budgets shows where the gap went. The size of what has to be split is not guessed either, it is added up in what the office costs the farm, and this sheet is where that total stops being one number and becomes a column.
What this sheet divides, and what it does not
It divides money the farm actually paid. It does not invent money the farm did not pay, and that second operation is a different one with a different name. The USDA cost of production series is explicit that the returns to owned resources cannot be established through market transactions during the production period and therefore have to be imputed, which is why they are estimated from a rate of return rather than divided from a receipt. Conab draws the same line by classifying the cost of opportunity as an implicit cost, one for which no actual disbursement occurs.
The practical reading for the sheet: the accountant’s fee is split, the rent you do not pay yourself on your own land is not. One is an invoice looking for two homes. The other is a figure somebody has to make up, defensibly, and mixing the two in the same column is how a sheet stops being checkable. Dividing what was paid is a control task, in the sense the four functions of management give the word, and valuing what was not paid is a planning one.
When the basis can be changed
When the change is written down with a date, and when the cycles on either side of it are not compared as if nothing happened. The USDA case is the model here, because that agency did change its basis: after the 1995 revision, general farm overhead moved from relative value of production to relative gross margins, and the documentation prints the old method and the new one in adjacent columns so a reader can see exactly what moved. That comparison is only possible because the earlier basis had been written down while it was still in use.
The same discipline is what makes the sheet worth keeping. The USDA handbook says it of its own accounts, writing about the leftover it reports as the return to management and risk: those figures are directly comparable only where the same procedure was applied to both of the products being set against each other. Two cycles of one farm are that same case with time in the place of the products.
The comparison says something when one basis produced both sides of it, and says nothing when it did not. File the split rule with the cost table it produced, not in a separate folder, and the next person who opens the table finds the ruler attached to it. That is the same habit as writing a declared price assumption beside each budget line, applied to a different document, and the finance axis is where both of them live.
Where to start
Two hours, with the list of shared costs on the table, on a day when nobody is waiting for an answer about any activity. The rule of the exercise is one line long: whoever writes the basis does not look at the result first.
A split rule that turns out to have been the wrong one, written down and dated, is worth more than the right one nobody recorded, because only the first can be argued with. The sheet is also cheapest to write on a day when nothing depends on it.
Once someone needs an answer about whether to shrink the cattle or drop the second crop, every basis on the list has a side, and the person choosing knows which side it is on. The three public bodies quoted here did not get past that by finding the correct basis. They got past it by fixing theirs in print years before any particular number became inconvenient, and by printing the day they changed it.