ERS puts labor at $1.85/cwt on farms above 2,000 cows and $13.18 on herds under 50 — while the big farms pay the higher wages. Almost none of the gap is pay. And $12.78 of it is you.

Rodney and Dorothy Elliott left a 140-cow farm in Northern Ireland to build Drumgoon Dairy near Lake Norden, South Dakota. Nearly two decades later, they were milking 6,500 cows with 20 robots and more than 50 employees. Some of those people had been with them since the earliest years.

In late May 2025, the Department of Homeland Security audited Drumgoon’s labor records. DHS determined that 38 workers had inaccurate, outdated, or incomplete citizenship or work-authorization documentation, according to South Dakota Searchlight’s October 2025 reporting. Elliott asked them for updated papers. Most couldn’t resolve the issues, and she had to let them go.

The crew went from more than 50 to 16 (South Dakota Searchlight, October 10, 2025; Northeast Radio SD, October 2025). Most of the 38 had worked at Drumgoon for years, the Searchlight reported, and some for nearly two decades — long enough to have had a hand in building the operation they were leaving. Elliott does not know where they went. Federal rules gave them ten business days from the audit finding to resolve their paperwork or be terminated.

Now the part that matters for anyone with a payroll. She didn’t post higher wages and wait for local applicants. She spent more than $110,000 on recruiters and transportation to bring 22 H-2A visa workers up from Mexico.

The Rebuild Nobody Talks About

This was an audit, not a raid. No agents in the yard, no arrests. Drumgoon had gone twenty years without one. Elliott told the Searchlight she reviewed applicants’ documents herself and had turned candidates away repeatedly over the years when the IDs looked questionable.

Metric	Before audit (May 2025)	After rebuild (Oct 2025)	What it cost
Crew size	50+ employees	38 workers	Still 10–15 short
Immediate post-audit crew	—	16 workers	6,500 cows on 16 people
Workforce composition	Local hires, some 20-year tenure	22 H-2A visa, 16 local/temp	Visa roles legally restricted
Workers terminated	—	38 (10 business days to cure)	Most had years of tenure
Recruitment and transport	$0	$110,000+	Recruiters and travel from Mexico
Wage increase posted	—	None reported	Money went to recruiting, not pay
Robot maintenance openings	Posted	Posted	Zero applicants

South Dakota Searchlight, October 10, 2025; Northeast Radio SD, October 2025. Elliott’s account describes recruitment and transportation spending; no wage increase was reported.

Note what the visa route couldn’t do. Those permits restricted which jobs the workers could legally perform, so Drumgoon still had to fill 16 positions locally. And she did all of this in mid-2025 — a full year before USCIS issued the memo that finally wrote down how dairy H-2A petitions get judged. The pathway existed. The clarity didn’t. Even rebuilt to 38, the farm sat short of where it started.

Sixteen people were now covering a 6,500-cow operation. Elliott told the Searchlight her remaining employees were making mistakes from the long hours, or because they were new to farm work — including backing a payloader into the manure pond. Some, she said, were getting only one or two days off in a 15-day period. Nearby farms sent workers over for a couple of days at a time through the summer.

“But what else do you do? Do you just let cows starve or calves die because there’s no one there to take care of them?”

— Dorothy Elliott, co-owner, Drumgoon Dairy, to South Dakota Searchlight, October 2025

Then the detail that should stop you cold. Drumgoon had 20 robots running before any of this happened, and posted maintenance positions aimed at graduates of the Lake Area Technical College program in the same county.

Twenty robots. A technical college down the road. Open skilled positions. And as of the October reporting, nobody had applied.

This wasn’t new either. Back in October 2023, two years before anyone audited anything, Drumgoon told a county zoning process it employed 22 people, 15 of them milkers and stall management operators, and that finding local workers was extremely difficult (Dakota Free Press, October 12, 2023). That’s on the record. The empty applicant pool isn’t a post-audit excuse — it’s a documented pre-existing condition.

Elliott put the question more directly than most operators would:

“We’ve achieved our goals we set out for ourselves: build a dairy, milk cows and grow the dairy industry in South Dakota. Is it a sustainable goal if there’s nobody to work on these dairies?”

Here’s the part that reframes all of it. ERS data says the largest dairies pay the highest wages in dairy — and still carry the lowest labor cost per hundredweight of any herd size class. A sub-50-cow herd shows $13.18/cwt. A 2,000-cow herd shows $1.85. If cheap labor built the big farms, that table should read the other way around.

Your $32 Billion Talking Point Is Older Than Your Replacement Heifers

Three numbers anchor nearly every dairy labor conversation in Washington and at every co-op annual meeting. Immigrant workers make up 51% of hired dairy labor. Farms employing them produce 79% of U.S. milk. Losing that workforce would cost the economy $32.1 billion.

The figure	Where it comes from
51% / 79% / $32.1B	Adcock, Anderson & Rosson, Texas A&M AgriLife Center for North American Studies
Published	September 2015
Data vintage	Survey fielded fall 2014; employment estimates for 2013
Funder	Commissioned by the National Milk Producers Federation
Independently replicated?	No — and the 2015 report was itself an update to a 2009 study by the same team for the same client

None of it is hidden. NMPF hosts the PDF and discloses the funding. But by 2024 and 2025, those numbers were circulating in trade coverage and House Agriculture Committee documents with no date attached.

Worth being precise about what “51%” actually measures, since this article is about dating your numbers. Researchers estimated 150,418 people worked on U.S. dairy farms in 2013, and that 76,968 of them — 51% — were immigrants. Thirteen-year-old employment data, restated in 2026 as though someone counted last week.

The same 2015 study found immigrant dairy employment rose 35%, nearly 20,000 people, over the six years since the 2009 survey — a number NMPF was still promoting in a December 2018 release. Funded by an industry association with a policy position, so weigh it accordingly. But it points up, not down.

If you cite 51/79 in a memo or at a hearing, attach the year. Costs nothing, and it makes you the most credible person in the room.

Why a Sub-50-Cow Herd Shows $13.18/cwt in Labor and a 2,000-Cow Herd Shows $1.85

USDA’s Economic Research Service breaks out dairy labor costs by herd size. Here’s the full table, including the column almost nobody quotes.

Herd size	Total labor/cwt	Unpaid family	Hired	Imputed wage for unpaid labor
10–49 cows	$13.18	$12.78	$0.40	$21.74/hr
50–99	$8.14	$7.53	$0.61	$22.18/hr
100–199	$5.12	$3.84	$1.28	$23.16/hr
200–499	$3.53	$1.45	$2.08	$23.71/hr
500–999	$2.87	$0.69	$2.18	$25.03/hr
1,000–1,999	$2.60	$0.30	$2.30	$25.09/hr
2,000+	$1.85	$0.10	$1.75	$25.81/hr

USDA ERS, ERR-274 (MacDonald et al., 2020), Appendix table A2, using ARMS 2016 Dairy Version, national. Hired column derived as total minus unpaid. The imputed-wage column values unpaid family labor at opportunity cost — it is not a hired pay rate.

That last column is the argument. The imputed wage rises steadily with herd size — $21.74 on the smallest farms to $25.81 on the largest. Hired wages do the same. ERR-274 doesn’t publish the hired rate by herd class, but MacDonald reports the same direction of travel: hired wage rates, like imputed ones, run higher on larger farms. That’s the whole point — the biggest dairies pay more per hour and still land at $1.85/cwt.

MacDonald states it flatly: differences in labor costs “do not arise from differences in hourly wage rates,” because wages for both hired and unpaid labor are higher on larger farms.

So the herd that shows $13.18/cwt isn’t paying more per hour. It’s paying $21.74 an hour of opportunity cost for a lot of hours spread across very little milk, and $12.78 of that $13.18 never touches a payroll cheque. It’s your hours and your family’s. Elliott’s 6,500 cows put her in the bottom row of that table. If you’re in the top one, you’re the unpaid labor line.

Look at the hired column too. It climbs to $2.30 at 1,000–1,999 cows and then drops to $1.75 on farms above 2,000 — despite those farms paying the highest wages in the table. That’s not a wage effect. That’s enough milk per worker to bury the cost.

Productivity gap, not wage gap. Which flips the question entirely. The 51/79 framing asks who’s milking the cows. The ERS data asks how many cows one person can milk.

The Expansion Math Was Never About Cheap Wages

We went looking for the counter-argument — that cheap labor was the precondition for building operations like Drumgoon. The expansion literature doesn’t support it.

Hadley, Wolf, and Harsh at Michigan State tracked 20 dairy farms through one-time herd increases of at least 20% between 1988 and 1998, and published their findings in the Journal of Dairy Science in 2002.

Measure	Preexpansion	Postexpansion
Herd size	296 cows	569 cows (+92%)
Milk per full-time equivalent	686,656 lb	917,980 lb (+34%)
Labor expense	$5.14/cwt	$3.50/cwt (−32%)
Debt-to-asset ratio	31.3%	43.4%

Hadley et al., Journal of Dairy Science 85(8), 2002. Debt-to-asset ratio reported for 14 of the 20 farms.

They didn’t get there by paying less. A 34% gain in milk per worker produced a 32% drop in labor cost per hundredweight — the same mechanism the ERS table shows, caught in real time during the buildout decade.

It wasn’t free either. Leverage went from 31.3% to 43.4% across the 14 farms reporting it. Same strategy, different landings.

Bewley, Palmer, and Jackson-Smith surveyed Wisconsin producers who modernized between 1994 and 1998, also in the Journal of Dairy Science, and asked what actually made expansion hard. Labor management ranked high. Wage rates didn’t make the list — and their finding that larger herds relied more on nonfamily labor while finding labor management easier is the whole argument in one sentence.

When the Labor Vanished, Nobody Got a Raise

Drumgoon isn’t the only case. In July 2025, at least nine Texas dairies received Notices of Inspection over a single weekend, Tyne Morgan reported for Dairy Herd Management on July 15. An NOI is a records request, not a finding of wrongdoing.

One farm and a Texas weekend are confirming evidence, not proof. Better to say so than let a handful of cases carry weight they can’t hold.

The stronger evidence sits outside dairy, and it’s causal.

The Bracero termination. The Johnson administration ended the program on December 31, 1964, excluding almost half a million Mexican seasonal farm workers. Clemens, Lewis and Postel studied it in the American Economic Reviewin 2018 and found no meaningful rise in domestic farm wages or employment. Growers mechanized instead — tomato harvesters went from a handful of units to near-universal inside about a year. The finding has a published critic: Kaestner argued in Econ Journal Watch in 2020 that the identification is weaker than claimed. It still stands as the best natural experiment available.

California’s AB 1066. The farmworker overtime phase-in began in January 2019 for employers with 26 or more workers, stepping the weekly threshold down from 55 hours to 40 by 2022. Alexandra Hill at UC Berkeley used National Agricultural Workers Survey data for 2019 and 2020 and found employers cut hours rather than pay premiums. The share working 56–60 hours a week — just under the old threshold — fell by roughly half. The share working 46–50 hours rose by about a third. Workers earned $6 to $9 million less in weekly paychecks across those two years, and the share earning $600–$800 a week dropped by roughly a third, most shifting into the $400–$500 bracket.

Different decades, different crops, different researchers. Both pointing where Drumgoon pointed. When labor gets scarce or expensive, employers reach for visas, machines, or fewer hours before they reach for a raise.

What Would Domestic-Only Labor Actually Cost You Per Hundredweight?

Fair warning on our own math first. The $1.75/cwt hired-labor figure is ARMS 2016, and the production cost is 2021. Two vintages in one equation, against a 2026 price. We just spent a section criticizing undated numbers, so it would be cheap not to date our own.

Here’s the calculation, and you can run the same shape of it on your own payroll in about four minutes. Take hired labor for 2,000-plus cow herds, $1.75/cwt, and apply a wage premium as though you’d replaced that workforce domestically.

Formula: $1.75 × your wage premium = added cost per cwt.

ERS puts 2021 total cost of production at $19.14/cwt for 2,000-plus cow herds. The 2026 all-milk price forecast has been sliding all summer: $20.70 in June, cut 70 cents to $20.00 on July 16, then cut another 15 cents in the August 24 outlook to $19.85/cwt.

Wage premium	Added cost/cwt	Total cost	Margin at $19.85
Baseline	—	$19.14	+$0.71
+20%	$0.35	$19.49	+$0.36
+40%	$0.70	$19.84	+$0.01
+60%	$1.05	$20.19	−$0.34

USDA ERS, Livestock, Dairy and Poultry Outlook, August 24, 2026. Every margin cell moves one-for-one with the milk price.

Watch what the August revision did. At a 40% wage premium, a 2,000-cow dairy now lands one cent above breakeven — it was sixteen cents in July. Thirty cents of forecast erosion did more damage to that row than a 20-point swing in the wage assumption.

Which is the actual finding. The worst-case labor shock costs a large herd about a dollar per hundredweight. The milk price moved 85 cents in ten weeks without anyone voting on it.

Now set both beside the herd size actually in trouble. ERS has sub-50-cow herds at $42.70/cwt in 2021 — $22.85 underwater against $19.85 milk, before labor enters the conversation at all. That’s the arithmetic closing barns, and it has nothing to do with immigration.

One more limit. We picked 20/40/60% as a sensitivity bracket because no study establishes what premium would actually pull domestic workers into dairy at scale. The model also assumes farms would pay it. Drumgoon, Bracero, and California all say they’d restructure or buy iron first. Elliott’s $110,000 went to recruiters, not a wage sheet.

Your Robot Breakeven Isn’t One Number. Salfer’s Own Range Runs $17.11 to $27.02

Bullvine has published the $27.05/hour breakeven repeatedly — and dated it to 2018 on at least one page. We went back to the source; what we found changes how you should use it.

That number comes from Jim Salfer and colleagues at the University of Minnesota, published in the Journal of Dairy Science in 2017, modeling a 1,500-cow dairy with 25 robots against a double-24 parlor. And it isn’t a single finding. It’s one cell in a sensitivity analysis.

Salfer’s 1,500-cow model, by assumption	Breakeven labor rate
1% wage inflation, robots give up 0.91 kg/d (about 2 lb)	$27.02/hour
3% wage inflation, equal production, 30-year horizon	$17.11/hour

Salfer et al., Journal of Dairy Science 100(9):7739–7749, 2017.

Read that again. Same researcher, same herd, same model — and the answer swings ten dollars an hour on two assumptions: whether your robots hold production, and what wages do over three decades.

We’ve been quoting only the top of that range. So has most of the industry.

Now the second land-grant number. UW-Madison Extension released its AMS Transition Budgeter on February 5, 2026, and the worked example runs 120 cows, two robots, a 5% milk bump, labor at $20.00/hour, boxes near $200,000 each. Breakeven wage: $14.77/hour. Since that farm already pays $20.00, the transition pencils. The tool’s rule is simple — if your actual labor cost is higher than the breakeven number, it works.

Here’s what nobody has connected. UW’s example is a 120-cow herd. And Salfer’s paper found robots penciling at 120 and 240 cows, while the 1,500-cow parlor beat the robots. Two land-grants, nine years apart, converging on the same range — and both saying scale cuts against automation, not for it.

So the apparent chasm between $14.77 and $27.02 was never a disagreement about robots. It was a disagreement about herd size, and about whether you assume production holds.

Our own Robot ROI Reality Check runs harder numbers than most quotes do:

Assumption	Dealer projection	Bullvine model
Installed cost	Dealer quote	1.4× dealer quote
Production gain	10–12%	6%
Maintenance	—	$11,500/robot/year
Downtime	—	6.5%

Bullvine modeling assumptions, not published research. Run your own quote through them.

On those inputs, a two-robot install on a 140-cow herd carries roughly an $8,776-a-year cash-flow hole for seven yearsbefore the math turns. Note the herd size — that figure is specific to a 140-cow model, not generic to any two-robot job.

Drumgoon is the sharper lesson anyway, and it isn’t the one in the brochures. Twenty robots didn’t stop that farm from losing 70% of its crew, and the skilled maintenance roles those robots created went unfilled. Automation changes what kind of labor you need — usually toward scarcer, better-paid labor. It doesn’t make you labor-proof.

Where the Evidence Still Runs Thin

Three honest gaps, because you’d spot them anyway.

Hadley’s cohort averaged 569 cows afterward — nowhere near Drumgoon’s 6,500. Whether the same productivity mechanism scales from 600 cows to 6,000 is an extrapolation, not a finding. And the debt-to-asset numbers come from 14 farms, not 20.

Salfer’s robot economics are from 2017, modeled on one 1,500-cow herd. Robot pricing, service contracts, and labor rates have all moved. The sensitivity logic holds; the dollar figures deserve a fresh run.

And nobody has done the direct study. No published work tests whether immigrant labor availability by region predicted where dairies expanded, holding feed cost, land price, and processing capacity constant. That’s the biggest hole in this entire debate, and it’s been sitting open for twenty years.

Options and Trade-Offs for Farmers

Path 1 — Run your own labor cost per hundredweight. Do this within 30 days.

When it makes sense: Any operation, any size; cheapest analysis here, and it tells you whether the rest of this applies to you.

What it requires: Annual payroll and annual hundredweight shipped. Divide one into the other. Work it on your own numbers — a 200-cow herd shipping 26,000 lb per cow moves 52,000 cwt a year, so a payroll of, say, $310,000 lands at $5.96/cwt. That example herd is deliberately labor-heavy. Swap in your two figures and see where you land against the $1.28 ERS reports for 100-to-199-cow herds and the $1.75 for 2,000-plus.

Risks and limits: Decide whether you’re valuing your own hours. ERS imputes $21.74/hour on sub-50-cow herds and $23.16 at 100–199. Skip that step, and you’re understating your real position — to yourself and to your lender.

Path 2 — Audit your I-9 files with counsel, also within 30 days.

The rules changed on March 16, 2026. ICE quietly updated its Form I-9 Inspection fact sheet, moving more than ten error categories from “technical” to “substantive.” Missing date of birth in Section 1. Missing date next to the employee signature. Incomplete List A, B, or C data in Section 2 — even where you kept document copies. Incomplete preparer or translator data. Electronic audit-trail deficiencies. Each now carries an immediate fine of $288 to $2,861 per formwith no cure period.

When it makes sense: Every operation with hired labor. No exceptions, and this is the risk that hasn’t priced in yet.

What it requires: Work with immigration counsel rather than alone. Ballard Spahr’s February 2026 guidance is blunt on the point — internal audits are what demonstrate good-faith compliance if a government audit lands. The statutory good-faith exception has always applied only to technical violations; that hasn’t changed. What changed is which errors count as technical. The ten-business-day cure window still exists for a shorter list: wrong Form I-9 version, missing “other last names used,” missing employee address in Section 1, missing business address in Section 2. Our full breakdown of how a Notice of Inspection unfolds walks through the mechanics step by step.

Risks and limits: You have three business days to produce every I-9 once a Notice of Inspection lands, and you must terminate workers with unresolvable documents within ten business days. Do the arithmetic on your own file count: 40 employees with one substantive error each, at the midpoint of that penalty range, is roughly $63,000 before anyone argues about aggravating factors. Drumgoon had a clean twenty-year record and a co-owner who personally checked IDs, and still lost 38 people. This reduces exposure. It doesn’t eliminate it.

Path 3 — Price your automation breakeven against your own herd size, not the brochure’s.

When it makes sense: Both land-grant models point at the same window. Salfer’s paper found robots penciling at 120 and 240 cows, while the parlor won at 1,500. UW’s worked example is 120 cows with a 5% milk bump and a $14.77 breakeven against $20.00 labor. If you’re between roughly 100 and 500 cows paying above $17/hour loaded, the math is live.

What it requires: A current dealer quote run at 1.4× installed cost, an honest production assumption, and a real budget line for maintenance skill. Drumgoon posted those positions and got nobody.

Risks and limits: The single biggest swing factor is production, not wage rate — that’s what moves Salfer’s breakeven from $17.11 to $27.02. Ask for the production guarantee in writing. And if a scale argument is doing the work in your automation decision, check it against the papers: both models put the economics at 120 to 240 cows, and Salfer’s parlor beat the robots at 1,500.

Path 4 — H-2A got clearer in June. Read what the memo actually says.

When it makes sense: Wider than early coverage suggested. On June 17, 2026, USCIS issued Policy Memorandum PM-602-0200, “Guidance on Temporary or Seasonal Need for H-2A Petitions for Dairying” — nine pages, effective immediately, binding on adjudicators. USDA welcomed it the same day. Per July 6, 2026 analysis, even dairies without a discrete breeding season may qualify by documenting materially different herdsman duties across the year, even though milking itself never stops.

What it requires: Documentation of seasonal duty variation, not of a labor shortage. And more lead time than you’d think — the contract, the certification, and a housing inspection all have to clear before anyone arrives, which puts realistic planning several months out.

Risks and limits: It’s a policy memorandum, not a regulation. It creates no legally enforceable right; any administration can rescind it, and petitions are judged case by case. Your year-round milking crew is still ineligible on its own. Elliott’s experience is the cautionary version: she was working this pathway in 2025, before the standard was written down, with counsel and $110,000 to spend — and it still left her 16 positions short. The Farm Workforce Modernization Act would put a year-round fix in statute, not a memo. It has passed the House twice and stalled in the Senate twice.

Key Takeaways

• Divide annual payroll by annual hundredweight shipped. Above 500 cows and well north of $1.75/cwt, your gap is labor efficiency, not your wage rate.
• Under 100 cows, value your own hours at ERS’s $21.74/hour before you call your cost of production finished. Otherwise, you’re the cheapest employee on the place, and nobody’s tracking it.
• If you audited your I-9 files before March 16, 2026, that audit is stale. Errors that were curable then now carry $288 to $2,861 per form with no correction window.
• Anyone quoting you one robot breakeven wage is quoting one cell of a sensitivity table. Ask which production assumption it uses. Equal production puts the bar near $17/hour; two pounds a day lost puts it near $27.
• Above 1,000 cows and weighing robots? Both Salfer and UW put the economics near 120 to 240 cows, and Salfer’s parlor beat the robots at 1,500.
• Modeling expansion? Track milk per FTE, not wage rate. The Michigan State cohort cut labor cost per cwt by 32% on productivity alone — and carried leverage from 31.3% to 43.4% getting there.
• Financing this year? Your lender’s labor-shock question has a bounded answer — $0.35 to $1.05/cwt on large herds. The milk price moved 85 cents against you in ten weeks. Know which one you’re actually exposed to.

So where does your labor cost per hundredweight actually sit — and how much of it are you paying versus quietly absorbing? Twenty minutes with your payroll file answers both, and it’s a better twenty minutes spent before an envelope arrives than after.

We’re running the complete scenario model — all seven ERS herd-size tiers, every premium cell, formula, and assumptions on the table — in an upcoming Bullvine deep dive. If you want the full math rather than the headline version, it’ll live there.

This article draws on reporting by South Dakota Searchlight (October 10, 2025), Northeast Radio SD (October 2025), and Dakota Free Press (October 12, 2023); on USDA ERS data and peer-reviewed research as cited; and on federal policy documents current as of September 2026. Drumgoon Dairy was not contacted for this article.

Learn More

• H-2A Dairy Visa Cost — Arms you with line-by-line guest worker recruitment expense data before signing agency contracts. Reveals true all-in costs averaging $877 per cow, exposing hidden legal overhead, transportation fees, and housing inspection mandates that conventional wage comparisons routinely conceal.
• Dairy Cost of Production: Small Herds — Dismantles the persistent myth that milk market consolidation is driven purely by feed volatility. Exposes how imputed family labor burdens of $12.78/cwt quietly suffocate sub-50-cow operations long before hired payroll changes impact the balance sheet.
• Robotic Milking ROI: Cash Flow Valley — Follows the money through a seven-year automated milking conversion to protect working capital. Breaks down why realistic 1.4× capital expenditure multipliers and $11,500 annual maintenance costs create an $8,776 yearly deficit on 140-cow setups despite brochure promises.

The Sunday Read Dairy Professionals Don’t Skip.

Every week, thousands of producers, breeders, and industry insiders open Bullvine Weekly for genetics insights, market shifts, and profit strategies they won’t find anywhere else. One email. Five minutes. Smarter decisions all week.