Diesel, Potash and the Next Farm-Cost Squeeze

A farmer buys fuel and fertilizer before selling a crop. That gap matters. Diesel prices have climbed during the conflict in the Middle East, while trade tensions make the future cost of Canadian potash harder to handicap. The first economic effect is a larger farm bill. Whether the farm products PPI and grocery prices follow is a separate question.

What is known, as of September 24: EIA reports U.S. on-highway diesel at $6.529 per gallon for September 21, up from $5.599 on August 31. Its September outlook expects constrained Middle East exports and low distillate inventories. USDA forecasts 2026 fuel and oil expense up 28.8% and fertilizer, lime and soil conditioner expense up 15.3% from 2025. EIA weekly price · EIA outlook · USDA forecast

Four prices, four points in the chain

The chart begins in February 2020, with each series set to 100. Nitrogen fertilizer and diesel are inputs; the broad farm products PPI tracks prices received upstream of food processing; the food CPI is what consumers pay. The different base units disappear, but the economic stages do not.

Source: RainbowStats retrieval of FRED series GASDESM, PCU325311325311, WPU01 and CPIUFDSL, indexed to the first displayed observation. Monthly observations through August 2026; food CPI has one fewer nonmissing observation. A shared index is for comparison, not a claim that each series responds equally.

The retail diesel series is a practical gauge of energy stress, although a farm's delivered fuel bill may differ. Nitrogen fertilizer is especially sensitive to energy inputs. Potash is a different nutrient: it supplies potassium, whereas PCU325311325311 tracks nitrogen fertilizer manufacturing. Treating the nitrogen coefficient as a potash tariff coefficient would be a category error. USGS explains the nutrient distinction.

The Canadian potash risk is conditional

Canada is an essential source of U.S. potash. But a threatened trade escalation is not the same as a new duty on every Canadian potash shipment. CBP described a 10% additional rate for Canadian potash outside USMCA preference in its 2025 tariff statement; qualifying shipments received different treatment. The September 2026 USTR action identifies targeted Canadian products and does not identify potash among its motor vehicle, dairy and alcohol targets. Trade policy could change, or border and sourcing frictions could raise delivered prices, but this article assumes no newly enacted across-the-board potash tariff. CBP · USTR

A proposed Belarus supply arrangement might push in the opposite direction, but its actual volume, freight cost and timing remain unsettled. The relevant policy variable is the delivered price farmers pay for potassium fertilizer, not a headline tariff rate alone.

What the lag analysis finds

I estimated two RainbowStats multivariate transfer functions using monthly log changes, allowing up to twelve monthly lags and retaining six terms. The first relates farm PPI shocks to diesel and nitrogen fertilizer shocks. The second relates food CPI shocks to farm PPI, nitrogen fertilizer and diesel. RainbowStats prewhitens the series with ARIMA, so the coefficients below describe relationships among ARIMA residual innovations.

Outcome	Selected terms (months after shock)	Observations	R²
Farm products PPI change	Diesel 0, 1, 6; nitrogen fertilizer 1, 2, 10	376	0.143
Food CPI change	Farm PPI 1, 3, 6; nitrogen fertilizer 0, 1, 9; no direct diesel term selected	376	0.132

Selected farm PPI transfer coefficients by lag. Positive early terms coexist with negative later terms. These are coefficients on prewhitened log-change shocks, not a projected percentage rise in farm prices.

In the farm PPI equation, diesel coefficients are +0.089 now, +0.117 after one month and −0.108 after six months. Nitrogen fertilizer coefficients are +0.137 and +0.164 after one and two months, then −0.144 at ten months. This pattern supports monitoring the first several months, but the later reversals and limited fit prevent a reliable cumulative forecast from this table alone. In the food equation, the selected farm PPI terms arrive at one, three and six months. Those associations give a plausible path toward the grocery shelf without proving that a particular diesel or tariff event caused it.

Six to nine months: a cost sensitivity, not a PPI forecast

USDA puts total 2026 farm production expenses at $492.8 billion. Its reported year-over-year changes imply roughly $21.5 billion in fuel and oil and $39.9 billion in fertilizer, lime and soil conditioners in its 2026 baseline (the estimates are reconstructed from rounded USDA changes). The following table applies additional, hypothetical price increases to those annual expense categories, holding quantities fixed. They are distinct from the increases USDA already forecasts.

Illustrative additional annualized spending above the USDA 2026 baseline. Scenarios are assumptions, not tariff schedules or econometric price forecasts.

Scenario	Fuel/oil price shock	Fertilizer category shock	Fuel cost	Fertilizer cost	Total added cost
Low	+10%	+5%	$2.15bn	$2.00bn	$4.14bn (0.84% of baseline expenses)
Middle	+20%	+10%	$4.29bn	$3.99bn	$8.29bn (1.68% of baseline expenses)
High	+30%	+20%	$6.44bn	$7.99bn	$14.43bn (2.93% of baseline expenses)

Arithmetic: additional fuel cost = $21.47bn × fuel shock; additional fertilizer category cost = $39.94bn × fertilizer shock. The fertilizer category includes more than potash, and a potash duty would affect only a subset of it. These rows are intentionally broad cost stresses, not estimates of what any proposed tariff would do.

From late September 2026, six to nine months means March through June 2027. Farmers may absorb a cost increase in margins, change application rates or crop mix, hedge or buy ahead; weather and global crop supply can dominate output prices. A spring cost squeeze is therefore plausible even if the broad farm PPI does not rise proportionately. EIA's September outlook assumes distillate inventories remain tight into 2027, while USDA's expense forecast already embeds large increases. The scenario table asks what happens if costs move still further.

The replication script

Paste these lines into RainbowStats. The first chart uses indexed levels; the two transfer models use monthly log changes. RainbowStats displays each submitted result, and the script is also included as a separate .rstats file in the download.

diesel=GASDESM
fert=PCU325311325311
farm=WPU01
food=CPIUFDSL
recentDiesel=since(diesel,20200101)
recentFert=since(fert,20200101)
recentFarm=since(farm,20200101)
recentFood=since(food,20200101)
set_title(line_chart(list(index_100(recentDiesel),index_100(recentFert),index_100(recentFarm),index_100(recentFood))),"Farm inputs, farm PPI and food CPI: February 2020 = 100")
farmChange=logdiff(farm)
fertChange=logdiff(fert)
dieselChange=logdiff(diesel)
foodChange=logdiff(food)
farmModel=MVTransfer_Function(list(farmChange,fertChange,dieselChange),12,6)
foodModel=MVTransfer_Function(list(foodChange,farmChange,fertChange,dieselChange),12,6)

Next improvement: add an actual farm-paid potash price or potassium fertilizer import-price series, then refit the farm equation with crop-specific outputs such as corn and wheat. A pre/post policy comparison should begin only after a duty is enacted and its coverage is known. Compare out-of-sample predictions and the 2020–22 episode before publishing numerical farm PPI forecasts.

Sources and scope

Model estimates were retrieved from RainbowStats on September 24, 2026. FRED series are generally revised; rerunning the script later may change the sample and selected terms. The model selects terms from many candidates, so reported t statistics are descriptive rather than clean post-selection tests. R² refers to fitted innovations, not to level-price forecasting performance.