Farm subsidy law begins with a promise that sounds narrow enough to test. The statutes that authorize federal farm support present the money as income support and stabilization for the farmers who produce the national food supply, delivered against a level of weather and price volatility that few other sectors match. A crop can be planted in a season of strong prices and harvested into a collapse that no decision on the farm caused, and a single drought, flood or freeze can erase a year of revenue across an entire region at once. The case written into the law is therefore not a case for generosity in the abstract. It is a case for a public instrument that steadies income when markets and weather move together against producers, keeps operations solvent through troughs they did not create, and preserves the capacity to plant the next crop. Whether the measured record shows the money landing on that aim is a separate question, and it is the question the evidence assembled here answers finding by finding.

The first difficulty in answering it is that the check and the benefit are not the same thing. A payment is made to a named recipient, recorded against a farm or a person, and reported in program statistics by the amount that left the Treasury. The economic benefit of that payment, however, travels. It can travel to the landowner who raises the rent once a reliable payment stream attaches to the land, to the seller who prices the expected stream into an acre offered for sale, to the lender whose loan becomes safer, or to consumers and competitors in other countries when support changes what is planted and at what price world markets clear. The record reviewed here follows the money past the farm gate and past the name on the payment, because the gap between the two is where the largest findings sit. Effects measured across the commodity, crop insurance and conservation titles of the recurring farm bills, in the form enacted by the Agriculture Improvement Act of 2018, Public Law 115-334, show a system that reaches the stated aim in part and misses it in part, and the pattern of the misses is consistent enough to name.

The organizing claim that the findings support is that federal farm support operates, in substantial part, as support for land, not for farmers. The distinction is not rhetorical. A payment keyed to acres, to historical production on those acres and to the value of what those acres produce will settle where the acres settle, in the ownership and rental structure of American farmland, and it will reward scale in proportion to scale. A payment that raises the return to owning farmland raises the price of farmland and the rent charged for its use, so a share of every dollar appropriated for operator income becomes a higher entry cost for the next operator. A payment designed to leave planting decisions untouched can approach that goal closely when it is genuinely detached from current prices and plantings, while a payment delivered through subsidized insurance changes the economics of marginal acres in ways the detached payment does not. Read together, the five findings describe a support system whose benefits concentrate, capitalize and leak in predictable directions, and whose clearest successes lie where the payment buys a defined outcome, such as retired fragile land, rather than where it supplements income in general.

The concentration finding gives the distribution its shape. Payment records assembled by the Environmental Working Group from United States Department of Agriculture data obtained under the Freedom of Information Act show the top tenth of recipients receiving more than three quarters of the subsidies recorded between 1995 and 2016, and the explanation travels with the figures: payments keyed to base acres, payment yields and production scale mechanically with the size of the operation, and production itself concentrated over the same decades. The capitalization finding then explains where a concentrated payment goes after it arrives, with tenant operators keeping about three quarters of a subsidy dollar in the rental research of Barrett E. Kirwan and landowners taking about a quarter through higher rents, against a tenure base in which roughly 39 percent of farmland in the contiguous states was rented and most rented acres belonged to landlords who did not farm. The production finding separates instruments that look alike in a budget table, finding decoupled payments close to neutral in their measured effect on planting while premium subsidies for crop insurance move marginal land into cropping. The cotton finding carries the argument abroad, recording that a trading partner priced the domestic system in a formal dispute and collected a settlement measured in hundreds of millions of dollars. The conservation finding closes the sequence with the most favorable evidence in the article, documented erosion, nutrient and habitat gains on retired land, paired with an open question about whether payments on working land purchase practices that would not have occurred without them.

The evidence behind those findings comes in distinct forms, and the distinctions matter for the weight each finding can bear. Payment records show distribution directly, since every dollar in the database was paid to a recorded recipient in a recorded year, and disagreement about the concentration figures concerns interpretation rather than the underlying records. Rental and land market studies work differently, using variation in policy and in payment rates to isolate the share of a subsidy that moves into rent or into price, so their results arrive as estimates with a direction firmer than their magnitude. Production studies face the hardest identification problem, separating the effect of a payment from the prices, weather and technology that move planting at the same time, and their convergence on small effects for decoupled payments across different data and methods is part of what the finding rests on. The cotton record is a formal legal record, findings issued by a panel and an Appellate Body and a settlement executed in signed agreements, while the conservation record mixes agency reporting on retired land with a research literature that treats working lands additionality as unresolved. An account that keeps those forms separate can say, finding by finding, what the record settles and what it leaves open.

What farm support actually is

Farm support, as federal law delivered it at the January 15, 2021 vantage, was not a single payment or a single idea about what government owed agriculture. Support arrived chiefly through three titles of the recurring farm bills, and each title moved money by a different route for a different stated reason. The commodity title provided price and income support. The crop insurance title subsidized premiums and the delivery of insurance against yield and revenue loss. The conservation title paid for land retirement and for conservation practices on land that remained in production. An operation could draw support through one channel, through two channels in the same year, or through all three, and the channels did not share the same limits, eligibility tests, or relationship to the crop grown in a given season.

The governing statute at that vantage was the Agriculture Improvement Act of 2018, Public Law 115-334, the most recent farm bill enacted by that date. Farm bills operated as recurring omnibus legislation that amended, extended, repealed, and refinanced programs carried in earlier bills, so support under the 2018 act descended through the Agricultural Act of 2014, Public Law 113-79, and beyond that act to the adjustment statutes of the New Deal era. Descent had practical force. Eligibility categories, acreage concepts, and payment formulas survived from one enactment to the next even when Congress changed program names or altered payment triggers. A producer dealing with the Farm Service Agency or a crop insurance agent under the 2018 act was therefore dealing with parts designed at different moments, for different market conditions, and fitted together rather than designed as a coherent whole at a single sitting.

The vocabulary of the system preserves that history. Base acres and payment yields look backward to a historical planting pattern and historical productivity, not to the farm as it stood in the year of payment. Reference prices look outward to market prices measured against levels fixed in statute. Marketing assistance loans tie support to a crop harvested and placed as collateral. Premium subsidies tie support to an insurance contract purchased for a specific crop year. Conservation payments tie support to land retired under contract or to practices maintained on working land. The anatomy is best understood as a set of distinct triggers, a price falling below a statutory level, county revenue falling below a benchmark, a premium coming due, or a contract obligating a landowner or operator to retire land or carry out a conservation practice. Different triggers produced different distributions, even when the dollars were reported under the single heading of farm support.

The commodity title formed the oldest channel. Under the Agricultural Act of 2014, commodity support rested principally on two programs with parallel purposes and different measures of loss, Agriculture Risk Coverage and Price Loss Coverage, known as ARC and PLC. Each paid on historical base acres and payment yields rather than on current plantings, a choice that separated payment from the planting decision made in the season in which the payment was earned. Beside ARC and PLC stood marketing assistance loans, an older form of support operating through credit secured by the harvested crop. Behind all three stood the direct payments paid from 1996 to 2013, which the 2014 act repealed.

How do ARC and PLC payments work?

ARC and PLC pay on historical base acres and payment yields rather than current plantings. PLC triggers when market prices fall below statutory reference prices. ARC triggers when county revenue falls below a benchmark built from prior prices and yields. Both programs were created by the Agricultural Act of 2014.

PLC addressed price alone. When the effective market price for a covered commodity fell below the reference price fixed for that commodity in statute, PLC generated a payment calculated from the gap between the two prices, applied to the farm’s payment yield and to a share of its base acres in that commodity. The reference price functioned as a statutory floor for measuring loss, not as a guaranteed price for the crop actually sold. A payment could therefore arise in a year when an individual farm’s yield was strong, because the trigger examined price rather than the revenue produced by price and yield together. PLC protected against a sustained decline in the market price of a covered commodity, and left yield risk largely to crop insurance.

ARC addressed revenue, the product of price and yield considered together. In its county form, ARC compared actual county revenue for a covered commodity with a benchmark revenue built from prior prices and prior yields, and a payment followed when actual county revenue fell below the guaranteed share of that benchmark. Payment thus depended on the experience of the county in which base acres were located, not solely on the individual operation holding those acres. A farm with a severe individual loss could receive a modest payment or none if county revenue held up, while a farm with a sound harvest could receive a payment if county revenue crossed the trigger. The county design traded farm level precision for administrative feasibility and for a loss measure reflecting conditions across a producing area.

The shared foundation was the base acre. Base acres represented historical acreage of covered commodities, established under earlier legislation and carried through the 2014 act, rather than acreage planted to those commodities in the payment year. Payment yields likewise represented historical productivity assigned to the base. Payment on that foundation is the feature described as decoupling, because support did not rise or fall with the acres planted to a covered commodity in the payment season and did not depend on the yield that planting produced. A producer could plant a different crop, plant beyond the base, or leave land idle without altering the base on which payment was calculated.

Decoupling was never absolute. Payments tied to historical base preserved a relationship between support and the scale and composition of past production, so operations with large historical plantings held large bases and stood to receive large payments when triggers were crossed. The expectation that a future farm bill might update base kept a further connection alive between planting decisions in one period and payments in a later period. The formula itself performed the distribution. Two facts therefore hold at once. In the payment year, ARC and PLC were separated from current plantings by design. Across farms, ARC and PLC remained scaled to historical production by the same design.

Marketing assistance loans preserved an older logic inside the commodity title. A producer of a covered commodity could obtain a loan using the harvested crop as collateral, at a loan rate established under the statute, and the structure provided cash at harvest without requiring an immediate sale into a harvest time market. If market prices recovered, the producer could repay the loan and sell the crop. If prices remained below the loan rate, the producer could settle on terms tied to a lower repayment rate or forfeit the collateral crop in satisfaction of the loan, so the loan rate functioned as a price floor for the crop actually produced and pledged. Unlike ARC and PLC, the loans were coupled to current production, because a crop had to exist, be eligible, and be placed under loan. Part of the benefit could appear as a marketing loan gain or a loan deficiency payment rather than as a payment calculated on base acres.

Direct payments supplied the clearest contrast. Authorized in the farm legislation of 1996 and paid through 2013, direct payments were made on base acres and payment yields at rates fixed in statute, without regard to market prices in the payment year and without regard to whether prices were high or low. A payment arrived in a prosperous year on the same basis as in a depressed year, because no price or revenue trigger stood between eligibility and payment. The Agricultural Act of 2014 repealed direct payments, together with the counter-cyclical payments that had accompanied them, and substituted the conditional logic of ARC and PLC, under which payment required an adverse price or revenue outcome measured against a reference price or a revenue benchmark. The repeal changed the timing and the public justification of payments. The continuity was equally real, because ARC and PLC retained base acres and payment yields, so the historical distribution of production continued to govern the distribution of support.

Crop insurance formed the second major channel, parallel to the commodity title rather than subordinate to it. Federal crop insurance, addressed in the crop insurance title of the farm bills, delivered support by subsidizing the premium paid for an insurance policy rather than by paying a producer directly after a measured loss. The Congressional Research Service, in report R43951, placed the average federal premium subsidy at about 62 percent of the premium, so the federal government paid the larger share of the policy cost and the producer paid the remainder. Policies insured yield or revenue for a specific crop, on specific acreage, in a specific year, which made insurance support coupled to current planting decisions in a way that payments on base acres were not. Planting more insured acreage, selecting higher coverage, or insuring a higher premium crop each raised the subsidy, because the subsidy was a share of a premium reflecting acreage, crop, coverage level, and production risk. The Agricultural Risk Protection Act of 2000 raised subsidy rates, and participation rose after that act to roughly 90 percent of eligible acres.

Delivery set the insurance channel apart as well. Commodity payments were calculated and paid through the Farm Service Agency from base acres, payment yields, and program elections recorded for a farm. Insurance subsidies flowed through the premium for a policy sold and serviced by private companies and agents operating under the federal program, with the Risk Management Agency setting the terms of offer and the federal government bearing premium subsidies, administrative support, and a share of underwriting risk. The substance did not depend on visibility, because a premium paid largely by the federal government lowered the cost of transferring yield and revenue risk, and the value of that transfer rose with the scale of the insured operation and the coverage selected.

The conservation title formed the third channel, and its payments purchased specified land management outcomes rather than compensating for price, yield, or revenue loss. The Conservation Reserve Program, created by the Food Security Act of 1985 and administered by the Farm Service Agency, paid annual rent under contracts running 10 to 15 years for retiring environmentally sensitive cropland and establishing protective cover. Enrollment removed land from crop production for the contract term, so land retirement had a direct relationship to acreage that payments on base acres had deliberately severed. Working lands programs paid on another basis. The Environmental Quality Incentives Program, dating to 1996, and the Conservation Stewardship Program, authorized in 2002 and renamed in 2008, provided cost share for conservation practices on land that remained in production, so federal money met part of the cost of a practice or management change on working land.

Payment limits exposed the sharpest structural asymmetry in the system, and the asymmetry lay between titles rather than within a single program. The Agricultural Act of 2014 set a per person payment limit of 125,000 dollars for ARC and PLC combined, so commodity payments to a single person in a year were capped at that combined level however large the base or however severe the triggering decline. The same act applied an adjusted gross income eligibility limit of 900,000 dollars to commodity program participation, so a person or entity above that level, measured under the prescribed rules, was ineligible for the payments the limit governed. Federal crop insurance premium subsidies carried no payment limit and no income means test. A large operation could insure acreage without a ceiling on total premium subsidy received, and a high income operation faced no adjusted gross income bar to premium support, even though commodity payments to the same operation were capped and income tested.

The asymmetry shaped behavior as well as distribution. An operation at the commodity limit gained no further ARC or PLC support from additional base, but gained further premium subsidy from insuring additional acreage or buying higher coverage, because no limit interrupted the link between insured production and subsidy. The asymmetry also defeats any simple statement about whether farm support as a whole was limited, because a statement accurate for the commodity title was inaccurate for the insurance title, and aggregate support to the largest operations depended on the mix between capped payments and uncapped subsidies in a given year. In a year of deep price decline, the commodity limit could bind hard while insurance support rose. In a year of strong prices, commodity payments could fall toward zero while premium subsidies continued at a scale set by insured acreage and coverage.

The stated aims of the statutes supplied the standard against which the anatomy has to be read. Across the farm bills that built the system, the commodity and insurance titles aimed to support and stabilize farm income against price and weather volatility. Prices for major field crops were set in national and international markets that no single producer influenced, and yields depended on weather, pests, and disease that no producer controlled, so income could fall sharply in a year of competent management and rise sharply in a year of ordinary management. A second stated aim was assurance of an abundant food supply, connecting income support to productive capacity on the reasoning that a producer base driven out by price troughs or catastrophic yield loss could not supply the capacity on which abundance depended in later years. The conservation title stated an environmental aim beside the income aims, to retire fragile land and to improve environmental performance on land in production.

The correspondence was imperfect in ways built into the formulas. Income stabilization pointed toward support that rose when income fell and receded when income was strong, and the trigger structure of ARC and PLC and the indemnity structure of crop insurance answered that description in form. Income support stated without a distributional qualification did not by itself specify whether support should follow historical production, current insured production, loss suffered, or doubt about an operation’s survival. The formulas answered by default. ARC and PLC answered with historical base and payment yields. Premium subsidies answered with insured acreage, crop value, and coverage level. Conservation payments allocated money by environmental criteria and contract terms rather than by income need. Whether that allocation served the stated aims is an empirical question about incidence and effects, not a question the statutory text resolves.

Scale was written into the anatomy before any administrator acted. Payments keyed to base acres, payment yields, and production scaled mechanically with operation size, and premium subsidies scaled with insured production for the same reason. A formula paying a rate per base acre pays ten times as much to the holder of ten times the base. A subsidy paying a share of premium pays more where the premium is larger, and premium is larger where more acres are insured at higher values against greater risk. Neither formula required an intention to concentrate support, and neither prevented concentration from following as arithmetic. Mechanical production of a result does not settle whether the result is defensible, because payment limits on the commodity title showed that Congress treated scale driven distribution as a matter for legislative choice, and the absence of limits on premium subsidies showed that the choice was made differently for a channel serving the same operations.

The channels also differed in the relationship each established between payment and land. Commodity payments attached to a farm’s historical record, while the land could be planted to other uses within statutory limits. Insurance subsidies attached to a crop in a field in a season and expired with that crop year’s policy, though the expectation of renewed subsidized coverage shaped decisions about which land to crop and which land to leave in grass or other cover. Conservation rents attached to specific enrolled acres for a decade or longer and restricted the use of those acres for the contract term. The differences mattered for the question of ultimate benefit, because a payment stream expected to continue could be reflected in the price or rent of the asset to which the stream was understood to attach, most clearly farmland.

The measurement question that the stated aims impose can be stated in precise form. If the purposes of the system were to support and stabilize farm income against price and weather volatility, to assure an abundant food supply, and to retire fragile land and improve environmental performance, then the evidence required is evidence about where the money actually landed, who ultimately captured its value after land markets adjusted, whether payments altered production decisions and on what land, whether support survived contact with international trade obligations, and whether conservation payments produced environmental results beyond what would have occurred without payment. Distribution follows from base, yield, premium, and scale. Capture follows from the attachment of payment streams to land that is rented and sold. Production effects follow from the degree to which payments separated from current planting remained separated in expectations and financial position. The anatomy does not prejudge those measurements, but measurement must proceed channel by channel and mechanism by mechanism, because a system built from different triggers, different limits, and different relationships to land and production could not produce a single uniform effect.

Finding One: A Payment Record Concentrated at the Top

The first question an impact record has to answer is who received the money. Farm subsidy law stated broad aims. Across the statutes, support was meant to support and stabilize farm income against price and weather volatility, and to assure an abundant food supply. Those aims describe a sector and a public purpose. The payment record describes a distribution. Between the stated aims and the recorded distribution sits the central fact of Finding One: federal farm payments flowed in large proportion to a small share of recipients, consistently enough across separate windows and scopes that distribution cannot be treated as a side detail. On the figures assembled from the government payment records themselves, concentration was the shape of the flow.

How concentrated were farm subsidy payments?

The Environmental Working Group Farm Subsidy Database, built from USDA payment records, found that between 1995 and 2016 the top 10 percent of recipients received 77 percent of recorded subsidies, because payments keyed to base acres, payment yields and production scale mechanically with operation size.

Each element of that answer matters, and each figure in the record needs its source, scope and years attached. The Environmental Working Group Farm Subsidy Database was built from USDA payment records obtained under the Freedom of Information Act. The records at its base were therefore government payment records, tracing payments the department itself had made. At the same time, the Environmental Working Group is an advocacy organization. Both facts belong in any account that uses the database, and neither cancels the other. The advocacy identity of the organization that assembled the database is a reason to read its framing with care and to keep scope and period attached to every figure. The Freedom of Information Act provenance of the underlying records is a reason to treat the tabulated payments as a record of government transactions rather than as an advocacy invention.

Stated with those cautions in place, the headline result is precise. Between 1995 and 2016, the top 10 percent of recipients received 77 percent of the subsidies recorded in the Environmental Working Group Farm Subsidy Database. The claim extends only as far as the recorded subsidies in the database extend, not to all federal spending and not to any single year. The longer window was not the only window in the record. In the Environmental Working Group release covering 1995 to 2010, the top 10 percent of subsidized farms collected 76 percent of all commodity payments. The scope there was commodity payments, a defined category within farm support, and the period was 1995 to 2010. Two measurements, over different periods and across different scopes, placed the share of the top tenth at roughly three quarters of the total in each case. Concentration of that order was not an artifact of one unusual year pulling a long average upward, nor of one unusually broad definition of support.

The commodity figures carry a detail that converts a share into a scale. For 1995 to 2010, the top 10 percent of subsidized farms collected 76 percent of all commodity payments, an average of 447,873 dollars per recipient across the 16 years. The average belongs to the same release, the same commodity payment scope and the same period. An average across 16 years is not an annual payment, and a reader who mistakes the 447,873 dollar figure for a yearly amount overstates the flow in any single year. Read correctly, the figure establishes cumulative scale: over a period covering several farm bills and several price cycles, the average recipient inside the top tenth accumulated commodity payments approaching half a million dollars, while the remaining nine tenths divided the remaining quarter of the commodity total among a far larger number of farms. In the Environmental Working Group figures for 1995 to 2010, the top 10 percent took 63 percent of commodity subsidies in 2010 alone. That share stood below the 76 percent share across the full period, showing variation rather than a fixed ratio, and the 2010 figure shows concentration at the end of the period as well, with one tenth of subsidized farms receiving nearly two thirds of commodity subsidies in that year alone.

Understanding why the mean and the lived center of the distribution diverge is essential to reading every figure in the record. A payment average is pulled toward the largest payments in a way that a count of recipients is not. When the top 10 percent of recipients received 77 percent of the subsidies recorded in the Environmental Working Group Farm Subsidy Database between 1995 and 2016, the average payment across all recipients described a flow dominated by the top of the distribution. The recipient in the middle stood in a part of the record where payments were far smaller. The statute promised income support in general language that treated farm income as a single subject, and the record delivered that support through a distribution in which the typical experience and the average figure pointed in different directions. An assessment that quotes an average without locating the median recipient misdescribes the program, because the average in a concentrated system describes the top carried across the whole, not the middle.

The mechanism behind the distribution was written into the payment formulas, and the production explanation deserves statement at full strength beside the figures. Payments were keyed to base acres, payment yields and production. A formula keyed to base acres pays in proportion to the historical acreage base attached to an operation, and a formula keyed to payment yields pays in proportion to the productivity figures assigned to that base. Such a system scales mechanically with operation size, because a larger operation commands more base acres, carries the yields attached to a larger base, and sits atop a larger production history than a smaller operation does. No discretion, favoritism or irregularity is required to produce a concentrated outcome from such a formula. The formula itself, applied evenly and exactly as written, produces concentration as its ordinary output, in the same way that a payment calculated as a rate multiplied by a quantity must return a larger total wherever the quantity is larger. A payment record keyed to the scale of production will mirror the scale of production, wherever production itself is concentrated.

Historical base deepened the scaling. Agriculture Risk Coverage and Price Loss Coverage, created by the Agricultural Act of 2014, were paid on historical base acres and payment yields rather than on current plantings. Payment therefore followed the recorded footprint of an operation, the base established under earlier program participation, more than the planting decisions of the season in which the payment arrived. An operation that had farmed extensively when base was established carried that footprint forward, while an operation with a small base carried a small footprint forward through the same formulas. Concentration could therefore persist through changes in prices and plantings, because the quantity that drove the payment changed far more slowly than those surrounding conditions.

Production itself was consolidating across the same decades, and the consolidation research supplies the second half of the production explanation. USDA Economic Research Service consolidation research by James MacDonald and coauthors found that farms with at least 1 million dollars in sales rose from 31 percent of the value of production in 1991 to about 51 percent in 2015. In 1991, farms at or above that threshold accounted for under one third of the value of production, and by 2015 farms at or above that threshold accounted for about one half. A payment system keyed to production, base acres and payment yields operated inside a sector whose output was moving steadily toward its largest operations, so even an unchanged formula would have delivered a rising share of payments to the largest farms. The Economic Research Service figures establish that the movement in production was real, measured independently of the subsidy database, and large enough to explain a substantial part of the payment concentration without any claim about manipulation of the payment system.

The land figures from the same research tell the consolidation story in physical terms. USDA Economic Research Service consolidation research by James MacDonald and coauthors found that midpoint cropland acreage roughly doubled between 1987 and 2012. Midpoint cropland acreage measures the acreage at the middle of the cropland distribution, the point at which half of all cropland sits on farms larger than the midpoint and half sits on farms smaller than the midpoint. A doubling of that midpoint between 1987 and 2012 means that the typical acre of cropland came to be located on an operation roughly twice the size of the operation that held the typical acre a generation earlier. A sector can retain a large count of small farms while most of its cropland, and therefore most of its production and most of its payment base, gathers into large operations.

One distinction must travel with the consolidation figures, because the distinction separates a documented fact from a familiar misreading. Roughly 98 percent of farms were family farms in the Economic Research Service record, so concentration by size is not corporate takeover. The largest operations in the production record, the farms with at least 1 million dollars in sales that rose from 31 percent of the value of production in 1991 to about 51 percent in 2015, were in their great majority family operations that had grown large, not outside corporations that had displaced family ownership. A family farm can operate at very large scale, farm a very large base and receive very large payments while remaining a family farm in ownership and management. The concentration finding raises questions about the distribution of public payments among farms of different sizes. The finding does not establish any claim that payments flowed mainly to corporate owners outside farm families, and an argument stated as a distribution among operations of different scale, most of them family owned, rests on figures the production and ownership records jointly support.

The distribution also sits inside a legislative structure larger than the commodity title alone. Farm support was delivered chiefly through three titles of the recurring farm bills, the commodity title, the crop insurance title and the conservation title, inside an omnibus measure whose reach extended well beyond commodity payments, a breadth captured in the observation that the farm bill’s commodity and nutrition titles draw on the same coalition and the same bill. A coalition assembled to pass an omnibus bill evaluates a title by whether the title holds the coalition together as well as by how the title distributes its own payments. Concentration inside the commodity title could therefore persist through successive farm bills without any single vote ever endorsing concentration as an aim. A formula that satisfied the largest producing regions and the largest operations carried coalition value independent of the distribution it produced. The result was a durable gap between the stated aims, which spoke of farm income in the aggregate, and the formulas, which allocated support in proportion to base, yield and scale.

What divides defenders of production based payments from advocates of stronger payment limits?

Program defenders stated the production explanation as a complete defense, and the defense has a coherent logic. On the defender reading, a payment system tied to production cannot distribute otherwise without ceasing to be production support. Payments keyed to base acres, payment yields and production exist to support the income generated by producing food at scale, and the operations that produce most of the food will, under any honest version of such a system, receive most of the payments. The farms with at least 1 million dollars in sales that accounted for about 51 percent of the value of production in 2015 could not plausibly receive only a small fraction of production support without the support detaching from production altogether. Program defenders pointed to the consolidation research of James MacDonald and coauthors as confirmation that the payment distribution tracked a real transformation in agriculture rather than a distortion imposed on agriculture by the payment system. On that reading, concentration in payments is the shadow cast by concentration in production, and a distribution weighted toward large operations is the expected signature of a program stabilizing the income of the sector that produces the food supply.

Payment limit advocates answered from within the same record and located the policy choice in a different place. On the advocate reading, payment limits with real force, and the absence of any limit in the insurance title, are design choices rather than necessities. A legislature that can write a formula keyed to base acres and payment yields can also write a ceiling into the formula, set the ceiling lower, apply the ceiling across titles and attach an income test to eligibility. Nothing in the arithmetic of production support compels a legislature to leave the largest operations free to accumulate payments in proportion to scale without bound. Payment limit advocates treated the production explanation as an accurate description of a mechanism and an inadequate justification of the mechanism, because the mechanism was itself chosen. The fact that payments scale mechanically with size explains how concentration happened, and the advocate position held that explanation and justification are different operations. A Congress dissatisfied with a distribution in which the top 10 percent of recipients received 77 percent of the subsidies recorded in the Environmental Working Group Farm Subsidy Database between 1995 and 2016 retained every legal tool needed to produce a different distribution. The dispute was about whether the recorded outcome, produced by chosen formulas under chosen limits, matched the aims the statutes themselves announced.

The text of the 2014 act gives each reading a concrete exhibit. The Agricultural Act of 2014 set a per person payment limit of 125,000 dollars for Agriculture Risk Coverage and Price Loss Coverage combined, and an adjusted gross income eligibility limit of 900,000 dollars. Limits therefore existed in the commodity title, written into the same statute that created the programs whose payments the limits governed. The 125,000 dollar limit represents a decision that some ceiling should stand between an individual recipient and the full proportional output of the formula, and the 900,000 dollar limit represents a second decision, that eligibility itself should end at a defined income level. Program defenders could point to those provisions as evidence that Congress had weighed distribution and set boundaries where Congress judged boundaries appropriate. Payment limit advocates could point to the same provisions as proof of concept: a statute capable of carrying both limits was capable of carrying stronger versions of either, and the level of each boundary was a judgment rather than a mathematical requirement of production support.

Against that commodity title structure stood the documented asymmetry of the crop insurance title. Federal crop insurance premium subsidies carried no payment limit and no income means test. The same farmer whose commodity payments were capped at 125,000 dollars per person for Agriculture Risk Coverage and Price Loss Coverage combined, and whose eligibility ended at an adjusted gross income of 900,000 dollars under the 2014 act, faced no corresponding ceiling on premium support through the insurance title and no corresponding income test on that support. Payment limit advocates treated that asymmetry as the clearest evidence that limits were design choices. A limit that governs one title and vanishes in another describes a legislative settlement reached title by title, rather than a necessity imposed by the subject matter. Program defenders answered that insurance operates on a different logic, that premium support purchases participation in a risk pool rather than delivering income in proportion to base, and that broad participation serves the stability of the insurance structure itself. The positions divide over what the same statutory facts prove: a calibrated settlement fitted to different instruments, or an inconsistency that reveals distribution as a choice made unevenly and capable of being made otherwise.

The stakes of the dispute come into focus against the statute’s stated aims, and the income stabilization aim is where concentration presses hardest. A program that stabilizes the income of a sector can be evaluated at the level of the sector total, and at that level concentration is almost irrelevant: a dollar of support stabilizes sector income by a dollar wherever in the distribution the dollar lands. A program that stabilizes the income of farm households reads differently. Household stabilization depends on how much support reaches the household in the middle of the distribution, the median recipient, whose payment bears little resemblance to the average payment in a system where the top 10 percent of recipients received 77 percent of recorded subsidies between 1995 and 2016. When the mean is carried by the largest operations, the mean describes the security of the largest operations more faithfully than the security of the typical recipient. The median recipient in such a system receives little relative to the program totals that descriptions of farm support usually quote, and the stabilization that recipient experiences is correspondingly modest. The statutory language never specified the level at which stabilization should be measured, and the concentration record forces the question the language left open.

The food supply aim cuts in the opposite direction and gives the defender reading its strongest ground. An abundant food supply is produced by output, and output was concentrating: farms with at least 1 million dollars in sales accounted for about 51 percent of the value of production in 2015, up from 31 percent in 1991 in the Economic Research Service research of James MacDonald and coauthors, while midpoint cropland acreage roughly doubled between 1987 and 2012. Support weighted toward the operations that produce about half the value of production is, by construction, support weighted toward the capacity on which the food supply aim depends. Stabilizing the income of large operations can therefore be defended as stabilizing the food supply in the most direct sense available to a payment program. The tension inside the statute is real: the income aim, read at the level of the typical farm household, is weakened by concentration, and the supply aim, read at the level of total output, is served by the same concentration.

The record assembled for Finding One therefore supports a layered account of what farm subsidy law actually did on distribution. At the level of measurement, the program delivered a concentrated flow: the top 10 percent of recipients received 77 percent of the subsidies recorded in the Environmental Working Group Farm Subsidy Database between 1995 and 2016; the top 10 percent of subsidized farms collected 76 percent of commodity payments between 1995 and 2010, an average of 447,873 dollars per recipient across the 16 years; and the top 10 percent took 63 percent of commodity subsidies in 2010 alone. At the level of mechanism, the flow was the predictable output of formulas keyed to base acres, payment yields and production, operating on a consolidating sector. At the level of ownership, the flow went overwhelmingly to family farms, roughly 98 percent of all farms in the Economic Research Service record, grown large rather than corporately owned. At the level of judgment, the record divides program defenders, for whom a production tied system cannot distribute otherwise and remain production support, from payment limit advocates, for whom limits of real force in the commodity title and any limit at all in the insurance title were available design choices. The measurement commands agreement, and the mechanism commands agreement. The judgment remains the live question the figures frame and cannot settle, because the figures describe a statute that promised income stabilization to a sector and delivered that stabilization in proportion to scale.

Finding Two: Capitalization and the Support That Settles in Land

A payment that arrives on a schedule, under rules that buyers and sellers expect to continue, does not remain a simple addition to the income of the person who receives the check in a given year. The payment attaches, through expectations, to the asset that qualifies for it, and in American agriculture that asset is land carrying base acres. A buyer who purchases land with an established base purchases two things at once, the productive capacity of the soil and the expected stream of government payments linked to that acreage, and the price paid reflects both. A tenant who rents land with base acres bids in a similar fashion, offering a rental rate that reflects not only what the acreage can earn from crops but also what the acreage can be expected to draw under the commodity programs. Over repeated sales and repeated rental negotiations, the expected stream becomes embedded in the value of the land itself, a process economists describe as capitalization. Once capitalization has occurred, a portion of every subsequent payment no longer rewards current production decisions at all. The portion instead services the price that was paid for the land, compensating the owner who sold or rented the acreage for surrendering a claim on payments that the acreage was expected to generate into the future.

The mechanism operates even when a payment is carefully designed to avoid distorting planting choices. A payment tied to historical base acres and payment yields rather than to current plantings, as Agriculture Risk Coverage and Price Loss Coverage payments were structured under the Agricultural Act of 2014, may leave a producer free to plant one crop rather than another without changing the payment received in that season. Freedom in the planting decision does not prevent capitalization, because the qualification for the payment runs with a defined land base rather than with a particular harvest. A prospective buyer can identify which parcels carry base, estimate the payments that base has generated, and incorporate that estimate into an offer. A prospective tenant can perform the same calculation before setting a rental bid. The decoupling of a payment from current production therefore separates the payment from planting incentives more successfully than it separates the payment from land markets. Land markets respond to any durable source of income associated with a parcel, whether that income originates in crop sales, in rental demand, or in a federal program with a history of renewal across successive farm bills.

The organizing finding of the capitalization evidence can be stated as a named claim: support for land, not for farmers. The claim does not assert that operators receive nothing, nor does it assert that every dollar of commodity support or premium support ends with a landowner. The claim asserts that a material and measurable share of support migrates away from the farming operation and into the price and rental value of the land, so that the ultimate beneficiary differs, in a substantial fraction of cases, from the beneficiary recorded on the payment. The claim matters because the stated aims carried across the farm statutes center on the income and stability of producers, on support against price and weather volatility, and on the preservation of an abundant food supply produced by an operating farm sector. A system that directs a quarter of a rental subsidy flow to owners who may not operate a farm, and that embeds a further share in purchase prices that new operators must finance, serves a different distribution than the distribution described by those stated aims, even when total outlays remain identical. The gap lies not in the size of the expenditure but in the incidence of the expenditure, in who finally gains after land markets have adjusted.

How much of a farm subsidy goes to the landowner instead of the farmer?

Barrett E. Kirwan’s estimates put the landowner share at about twenty five percent, with rental rates rising by roughly twenty five cents for each subsidy dollar, while tenant farmers retained about seventy five percent of the payment despite theory predicting that owners would capture nearly all of it.

The full study behind those figures is Barrett E. Kirwan, “The Incidence of U.S. Agricultural Subsidies on Farmland Rental Rates,” Journal of Political Economy, 2009. Kirwan examined the incidence question directly, asking how a subsidy paid in connection with rented farmland divides between the tenant who cultivates the land and the owner who supplies the land, once rental rates have had an opportunity to respond. The theoretical expectation entering the analysis assigned nearly the whole subsidy to landowners. That expectation rested on a familiar argument about fixed supply. Land suitable for a program crop in a given locality exists in a quantity that cannot expand quickly in response to a payment, while tenants can compete against one another for access to that land. Competition among tenants for a scarce and payment bearing asset should, under the simple version of the argument, bid rental rates upward until the advantage of the subsidy has been transferred to the owner, leaving the tenant in approximately the same economic position occupied before the payment was introduced. Under that logic, raising the subsidy would raise rents by a comparable amount, and a policy described as income support for producers would function in practice as a transfer to property owners.

Kirwan’s quasi-experimental estimates corrected that expectation without overturning its central warning. Tenant farmers, the estimates found, capture about 75 percent of the subsidy, and landowners capture about 25 percent through higher rental rates, with rental rates rising by roughly 25 cents for each dollar of subsidy. The correction is substantial. A tenant who rents program land retains three quarters of the associated payment rather than surrendering the whole amount at the next lease negotiation, a result that Kirwan attributed to imperfections in rental markets that limit the ability of owners to extract the full value of a payment through rent alone. Contracting frictions, local relationships, incomplete information, and the practical limits on rent adjustment within established tenancies all slow the transfer that the simplest theory predicted would proceed to completion. At the same time, the confirmation inside the correction is equally substantial. One dollar in four does leak to the landowner through the rental channel, year after year, on rented acreage across the program area, and the aggregate size of that leak grows mechanically with the aggregate size of the payments. A quarter share applied to a large national program is not a rounding error at the margin of the system. The share represents a persistent redirection of support, built into the ordinary operation of rental markets rather than produced by fraud, error, or evasion of program rules.

A further detail in Kirwan’s findings sharpens the interpretation. The landlord share rises where rental markets are more competitive. Where more tenants compete openly for available land and rents adjust more freely to changing returns, owners capture a larger fraction of the subsidy, moving the observed outcome closer to the theoretical prediction. Where markets operate with greater friction, tenants retain more, not because the program design protects the tenant share but because the market mechanism that transfers value to owners functions less completely. Protection that arises from friction is not the same as protection written into statute. A change in contracting practice or in the information available to owners could enlarge the landowner share without any amendment to the farm legislation itself. The incidence of the subsidy is therefore determined jointly by Congress, which sets the payment, and by land markets, which redistribute the payment after the payment has been authorized.

Rental rates capture only one channel of capitalization. The second channel operates through sale prices, where an expected payment stream raises the capital value of the land itself. Research published in 2003 examined that channel from complementary directions. Goodwin, Mishra and Ortalo-Magné, in work published in 2003, found that decoupled payments are capitalized into farmland values, and Roberts, Kirwan and Hopkins, in work also published in 2003, reached findings on the capitalization of payments into farmland values that pointed in the same direction. The convergence of the two lines of research carries weight because the rental evidence and the value evidence describe the same economic process observed at different points. A payment that raises rental rates by a measurable amount each year raises the amount a buyer will pay for the right to collect those rents in the future, because the purchase price of an income producing asset reflects the capitalized value of its expected income. A higher rent and a higher sale price are therefore not two separate effects competing for explanation. Each effect implies the other, and evidence for capitalization in rental markets predicts capitalization in sales markets, just as evidence for capitalization in sales markets confirms that buyers treated the payment stream as a durable attribute of the land being purchased.

The magnitude of the value effect received an official estimate in the same period. The 2003 Report of the Commission on the Application of Payment Limitations for Agriculture estimated that government payments had raised United States farmland values by 15 to 25 percent. The estimate, attributed here to that commission and to that year, describes the aggregate consequence of capitalization across the farmland base. An increase of that scale means that farmland prices in the United States during the period examined carried a substantial component that reflected expected federal support rather than expected earnings from agricultural production alone. A parcel that would have traded at a given price on the strength of its soils, location, and crop returns traded instead at a price elevated by one sixth to one quarter once the payment stream was incorporated. The commission estimate also converts the incidence debate from a discussion of annual flows into a discussion of national wealth. Annual rental leakage describes who receives a payment in a particular year. A capitalized value increase of 15 to 25 percent describes a one time and then continuing revaluation of the largest asset class in the farm sector, conferring a gain on whoever owned farmland when expectations of continued payments became embedded in prices, and imposing a corresponding cost on whoever purchased farmland afterward.

Understanding where the capitalized share flows requires a clear account of who owns the rented farmland base, and the most complete account available at the January 2021 vantage came from the Tenure, Ownership, and Transition of Agricultural Land survey, known as the TOTAL survey, conducted by the United States Department of Agriculture Economic Research Service and the National Agricultural Statistics Service in 2014. The survey found that approximately 39 percent of farmland in the contiguous 48 states was rented, about 354 million acres out of 911 million acres of farmland in that geography. The rented share of cropland was higher than the rented share of farmland as a whole, with more than half of cropland rented. Rental is therefore not a peripheral arrangement at the edge of American agriculture. For the crop sector that receives the bulk of commodity program support, renting represents the dominant tenure arrangement on a majority of the acreage involved. Any incidence effect that operates through rental markets, including the 25 percent landowner share identified by Kirwan, applies across that enormous base rather than across a small class of exceptional tenancies. The scale of the rented base transforms a proportional finding into a structural feature of the support system.

The ownership composition of the rented base extends the implication further. The 2014 TOTAL survey found that 80 percent of rented acres, 283 million acres, was owned by non-operator landlords, meaning owners who do not farm the land themselves. Retired farmers made up 38 percent of non-operator landlords, the largest single category identified within that ownership group. The remaining non-operator ownership includes heirs who inherited farmland without taking up farming, investors, and other owners whose connection to the acreage runs through title rather than through daily operation. A large share of the rented base is therefore held by people who do not farm, including retired farmers and heirs, so the capitalized share flows substantially outside the operating farm sector. The flow to retired farmers deserves precise characterization rather than dismissal. A retired farmer who rents land to a successor receives, through the capitalized portion of rent, a form of retirement income partly financed by a program whose payments are recorded as support for the current operator. That income may be entirely legitimate in human terms and may reward a lifetime of work on the same acreage, but the economic classification remains unchanged. Support recorded against current production is financing, in part, the retirement of a previous generation of producers and the inheritances of owners who never farmed the land at all.

The divergence between beneficiary of record and economic beneficiary follows directly from the combination of the rental evidence and the tenure evidence. Program administration records a payment in connection with an operation, and public discussion of farm support generally describes that operation, and the family behind that operation, as the recipient of public assistance. Economic incidence, after land markets adjust, tells a more complicated story in which approximately a quarter of the value on rented land settles with the owner, and in which elevated land values transfer a further portion of the program benefit to sellers at the moment of sale. A landowner who captures a share through a higher rent receives no check from the Farm Service Agency and appears in no recipient database, yet the gain is real and recurs for as long as the tenancy and the program continue together. A landowner who sells at a capitalized price captures the expected value of many future years of payments in a single transaction, converting an annual public expenditure into a private capital gain at the point of exit from farming or from land ownership. Accounting that follows only the check therefore understates the degree to which the support system has become a system for sustaining land values, and overstates the degree to which each appropriated dollar strengthens the competitive position of the person actively farming the land in that year.

The entry barrier created by capitalized prices may be the most consequential effect of all, because the barrier falls precisely on the producers whom farm policy rhetoric most often promises to assist. A beginning farmer who purchases land must pay the capitalized price, which includes the full market value of the expected subsidy stream attached to the base acres on that land. Having paid for the stream at purchase, financed in most cases through borrowed money carrying interest, the beginning farmer then receives the annual payments less as a net addition to income and more as a partial reimbursement for a cost already incurred. The payment helps service the debt contracted to buy the payment. The net support left over after that service is far smaller than the nominal payment figure suggests, and in a fully capitalized purchase the net support approaches the small residual that markets failed to price. The structure produces a paradox at the heart of the support system. The longer and more reliably a program operates, the more completely its value becomes priced into land, and the less assistance the program provides to anyone entering farming after that pricing has occurred. Generous support in one decade becomes an elevated cost of entry in the next, so that the program simultaneously assists established owners and taxes, through land prices, the successors whose entry the same program ostensibly encourages.

The intergenerational transfer embedded in land prices operates by the same logic across a longer horizon. When payments become capitalized, the benefit of a program concentrates in the generation that owned land at the moment expectations changed, because that generation receives both the annual payments during ownership and the elevated price at sale. Later generations receive the payments as well, but later generations pay for the privilege through higher purchase prices and higher rents, surrendering much of the benefit back to the preceding owners. A program renewed across successive farm bills from one decade to the next therefore does not distribute its benefits evenly across the successive generations who farm under that program. The distribution tilts toward earlier owners and away from later entrants, toward exit and away from entry, and toward inheritance of appreciated land and away from purchase of land at prices unassisted by public payments. Heirs who inherit capitalized land receive the embedded value without having paid for the stream, which softens the barrier for transfer within a family while leaving the barrier at full height for a beginning farmer purchasing outside a family succession. The support system, measured across generations, functions substantially as a transfer to the asset position of one cohort of owners, financed by taxpayers and repaid in part by the next cohort of operators through the land market itself.

Capitalization also creates a powerful source of political durability for the programs involved, independent of any judgment about the merits of farm income support in the abstract. Once a payment stream has been capitalized into land values, termination or sharp reduction of the program no longer merely reduces the annual income of current operators. Termination imposes capital losses on landowners who paid for the expected stream, whether those owners are retired farmers holding rental land, heirs holding inherited acreage, or operators who purchased land at prices that assumed continued support. A landowner who bought acreage at a price elevated by 15 to 25 percent in the estimation of the 2003 Commission on the Application of Payment Limitations for Agriculture faces a loss of that portion of the investment if the expected payments disappear. Lenders who advanced money against capitalized values face a corresponding deterioration in the collateral behind farm real estate loans. Opposition to program reduction consequently assembles itself from a constituency far broader than the operators who receive payments in any single year, and the opposition carries the sympathetic claim of reliance, the argument that a purchase made in good faith on the basis of an established public program should not be devalued by a sudden change in that program. That reliance claim grows stronger with each passing year of program continuity, because each sale at a capitalized price creates a new owner with a fresh reliance interest, which means that time itself, under a capitalized system, manufactures the political constituency for the continuation of the system.

A final consequence concerns the accuracy of public debate over the size and purpose of farm support. Debate framed around payment records treats the operator named in those records as the beneficiary and treats the payment amount as the measure of assistance received. Capitalization shows both premises to be incomplete. The named operator on rented land retains, on Kirwan’s estimates, about 75 percent of the payment, a majority share that should not be minimized, while about 25 percent moves to an owner whose name appears in property records rather than in program records. The purchasing operator retains even less in net terms after paying a capitalized price for the land. The buyer of land, the tenant in a competitive rental market, and the heir outside a family succession each encounter the support system in a different form, as cost, as partial transfer, or as windfall, and any assessment of what farm subsidy law accomplished that ignores those differences confuses the disbursement of money with the delivery of benefit. The central claim stated at the outset survives the qualifications that the evidence requires. The tenant share is larger than theory predicted, the landowner share is a quarter rather than the whole, and the value effect is an estimate rather than a measurement of any individual parcel. After those qualifications have been granted in full, the remaining fact stands at the center of the record. A substantial share of American farm support was converted, through the ordinary operation of land and rental markets, into support for land itself, accruing to owners at sale and at rent negotiation, raising the price of entry for the next generation, and binding the continuation of the programs to the protection of capitalized wealth long after the payments that created that wealth had been spent.

Finding Three: Production Changed at the Margins

The central design question in farm subsidy law was whether a payment changed what a producer planted and how much a producer grew. That question separated income support from production subsidy in economic substance, even when both arrived as a federal check drawn on the same appropriation. A payment that left planting decisions untouched transferred income without instructing the market to produce more of one crop, less of another, or any crop at all on land that would otherwise have rested in grass, fallow or another use. A payment that moved acreage, input use or output did more than support a household. It altered supply, pressed on prices, pulled land between uses and carried consequences for trade partners and for the soil, water and habitat attached to the land whose use had changed.

Decoupling was the central design project of the reforms enacted in 1996 because it promised to break the mechanical link between support and production. Under coupled support, the reward rose with the crop grown, the price recorded or the acres planted in a particular commodity, and a rational producer read the program itself as part of the price of expanding output. Under the decoupled design that followed, the 1996 act’s decoupling experiment replaced price linked support with fixed payments on historical acres, and the payment arrived according to a history the producer could no longer change by planting differently in the season ahead. The acre that had qualified in the base period carried the payment whether its operator seeded it, left it idle within program rules or turned it to a different crop. On paper, the marginal decision faced in the field had been freed from the subsidy, and the producer was to plant for the market alone.

The literature on production effects began from a refusal to accept that paper description as the whole answer. A payment severed from planting in its formula could reach planting through the balance sheet, through the psychology of risk, through the constraints on borrowed money and hired labor, and through expectations about what a future Congress might reward. Each channel operated without any rule requiring a producer to grow a particular crop, and each converted a fixed transfer into a change in behavior at the edge of a decision. The task taken up by agricultural economists after 1996 was to identify those channels, to measure the response each one produced, and to compare the total against the response that coupled payments had produced dollar for dollar. The result was consistent in direction and humbling in scale for both camps in the debate. Decoupled payments did affect production, so the claim of complete neutrality failed. The effects were small, so the claim that a decoupled payment functioned as a disguised version of the coupled payment it replaced also failed.

The wealth and insurance effects supplied the theoretical foundation, and David A. Hennessy articulated both in the American Journal of Agricultural Economics in 1998. The wealth effect rested on a proposition about how producers bear risk. A producer without wealth faces every uncertain season with a limited capacity to absorb loss, and prudence under scarcity takes the form of caution in planting. Crops differ in the spread of their possible returns as well as in their average return, and a cautious producer favors the safer pattern even when a riskier pattern promises more income on average. A government payment that raised wealth changed that calculation without changing any price. As wealth rose, the absolute amount of risk a producer was willing to bear rose with it when preferences displayed decreasing absolute risk aversion, meaning that richer decision makers accept larger absolute gambles than poorer decision makers facing the same proportional stakes. The payment therefore nudged the producer toward the riskier and often more productive planting pattern that caution had previously suppressed.

Hennessy developed the companion insurance effect from the stabilizing character of a predictable payment. A fixed payment whose arrival did not depend on the harvest reduced the variance of total farm income across good and bad seasons, and that reduction functioned in part as insurance even though no policy had been issued and no loss had been indemnified. A producer whose income floor had been raised could commit acres, seed, fertilizer and machinery time to a plan whose returns fluctuated, because the downside of a poor outcome had been softened by income that arrived regardless. The two effects pointed in the same direction and were difficult to separate in observed data, since a payment that made a producer wealthier also made income steadier when it recurred on a known schedule. Both depended on risk aversion as the transmission belt. A producer indifferent to risk had no caution for wealth to relax and no variance for stability to offset, and for such a producer a truly fixed payment remained neutral. The theory therefore predicted effects that were real among risk averse operators, uneven across farms, and bounded by the size of the payment relative to the wealth and income already at stake.

Credit and liquidity relief formed a second channel, grounded in the calendar of farm finance rather than in preferences alone. Production required spending before revenue arrived. Seed, fuel, fertilizer, repairs and rent fell due months before a harvest could be sold, and many operations bridged that gap with borrowed operating money whose availability depended on collateral, repayment history and a lender assessment of the season ahead. A producer constrained by credit did not always plant the pattern that expected market returns favored. The constrained producer planted what could be financed, deferred inputs whose payoff was probable but distant, or left productive capacity unused because a lender would not extend the line far enough to employ it. A decoupled payment relaxed that constraint in two ways. Cash in hand reduced the amount that had to be borrowed, and a predictable stream of future payments improved the repayment picture a lender examined when deciding how much to advance. The payment could therefore expand production not by raising the return to an additional acre but by making the additional acre financeable.

The credit channel explained why identical payments produced different responses on different farms. An operation with ample reserves gained liquidity it did not need, and its planting pattern had little reason to move. An operation at the edge of its credit limit gained room to act on plans that had been ready in every respect except financing. Lenders also treated expected support as part of the environment in which loans were made, so a dependable payment stream could soften the contraction of credit in a poor price season and keep acres planted through a trough, which was a production effect even though no payment formula had instructed the planting.

Labor supply supplied a third channel, and its direction was less certain. Farm households supplied labor to their own operations and to off farm employment, and a payment that raised household income without requiring additional farm work changed that allocation. Additional income could reduce the pressure to seek off farm wages, freeing household time for more intensive management or for retaining acres that a labor shortage might have forced out of use. The same income could also purchase leisure or finance hired labor that substituted for household effort. The labor channel therefore connected payments to production through the household rather than through the field, and empirical work had to allow for responses that offset one another inside the same population of recipients.

Expectations supplied a fourth channel, and it proved the most damaging to the claim that the post 1996 payments had achieved full decoupling. A payment formula written as fixed in one statute existed inside a legislative system that rewrote farm law on a recurring cycle, and producers understood that the base on which a payment rested might be recalculated in a future act. If planting more acres of a program crop in the present raised the chance of a larger base in the future, then the present planting decision carried a shadow return beyond the market price of the crop itself. The producer was planting partly for the market and partly for position in the next allocation. Daniel A. Sumner made that critique in 2003 in connection with the farm legislation enacted in 2002, which had permitted producers to update base acres. The offer to update converted a theoretical possibility into a demonstrated practice, and a rational producer thereafter treated the possibility of another revision as part of the return to planting program crops.

The expectations channel differed from the others in an important respect. Wealth, insurance, credit and labor effects operated even under a payment that every producer believed would never be recalculated, because those effects flowed from the possession of money and security rather than from the formula. The expectations effect operated only because the formula might change, and its size depended on beliefs about legislative behavior that no model could observe directly. Sumner’s argument implied that formally decoupled payments were never fully decoupled in expectation, since the legislative process itself supplied the coupling that the statutory text had removed. Any government that reserved the right to revisit a historical allocation taught recipients to farm the allocation rules alongside the market.

Do decoupled farm payments increase production?

Evidence from studies published between 1998 and 2012 found that decoupled payments raised production by a small amount through wealth, credit and risk channels, but the measured acreage and output responses were far smaller than the responses to payments tied directly to current prices or plantings.

Barry K. Goodwin and Ashok K. Mishra examined the acreage and resource allocation effects of decoupled payments in work published in 2006 and found only modest responses. Their result acquired much of its force from a simple comparison of scale. Payments in their data averaged only about 1.8 percent of the net worth of the average Corn Belt farm, and a transfer of that relative size could relax risk aversion or ease a credit limit only so far. The wealth effect predicted by Hennessy depended on the payment moving the producer wealth position enough to change the absolute risk the producer was willing to carry, and a payment amounting to a small fraction of net worth moved that position by a correspondingly small amount. The Goodwin and Mishra finding therefore did not refute the theory. It disciplined the theory with arithmetic.

Jeremy G. Weber and Nigel Key, writing in the American Journal of Agricultural Economics in 2012, found small positive effects of direct payments on cropland acreage and traced the response through the wealth and credit channels that theory had identified. Their work mattered because it located the effect where the mechanism said the effect should appear. Payments did not raise acreage uniformly or by an amount suggesting that producers treated each dollar as a planting instruction. The response ran through the financial position of the operation, which meant that the balance sheet mediated between the transfer and the field. That pattern supported the interpretation that decoupled payments influenced production as a byproduct of improved financial capacity rather than as a bounty on output, and it implied that the effect would concentrate among operations whose financial constraints had been binding before the payment arrived.

Teresa Serra, Barry K. Goodwin and Allen M. Featherstone added evidence on the risk mechanism itself in work published in 2011. They found empirical support for decreasing absolute risk aversion among producers, confirming the behavioral premise on which the wealth effect rested, and at the same time estimated a very small elasticity of output with respect to decoupled payments. The pairing captured the mature position of the literature with unusual clarity. The premise was true and the consequence was tiny. Producers did become less absolutely risk averse as wealth rose, and output did respond in the predicted direction, but the proportionate change in output following a proportionate change in payments was small enough that the payment remained principally an income instrument.

Comparative work by the Organisation for Economic Co-operation and Development placed the American findings in an international frame and reached a consistent ranking. Across the major support instruments in use among member countries, decoupled payments ranked as the least production distorting, and the output effect associated with a dollar of decoupled support stood far below the output effect associated with a dollar of market price support or a dollar of payments linked to output. The ranking followed from the structure of the instruments. Market price support raised the return on every unit produced and rewarded expansion directly. A decoupled payment reached production only after passing through wealth, risk, credit, labor or expectations, and each passage diluted its force. Support instruments could be arranged along a spectrum of distortion, and moving a dollar from a coupled instrument toward a decoupled instrument reduced the production effect of that dollar even though the reduction never reached zero.

The verdict the combined literature supported was therefore narrow and specific. Decoupled payments had real production effects that operated through identified channels and appeared in careful empirical work. Those effects were small in magnitude, far below the effects of coupled payments, and small enough that the principal economic action of a decoupled payment remained the transfer of income and the stabilization of the recipient household. Both halves of the verdict carried policy weight. The first half ended the strongest form of the decoupling claim, under which a fixed payment was treated as production neutral for purposes of trade discipline and environmental assessment. The second half preserved the reform achievement of 1996 against the charge that nothing had changed, since an instrument whose distortion had fallen so far represented a genuine change in the relationship between the Treasury and the field.

The contrast with federal crop insurance sharpened every part of that verdict, because the insurance title produced measured production effects larger than those attributed to decoupled payments, and it produced them through a more direct alteration of the returns to planting risky land. Where the crop insurance program’s statutory structure and delivery system channeled support through premium subsidies rather than through fixed payments on historical acres, the subsidy attached to a decision made in the present season about land in the present season. A producer deciding whether to crop a particular field compared the expected return of cropping, including the protection an insured revenue or yield outcome provided, against the return of leaving the field in grass or another use. Lowering the producer cost of that protection raised the relative return of cropping at exactly the margin where land use was decided. The insurance subsidy therefore did not need to travel through wealth or expectations to reach planting. It entered the calculation that determined planting.

The scale of the premium subsidy explained its reach. According to the Congressional Research Service, the federal premium subsidy averaged about 62 percent of the premium, so the producer paid a minority share of the cost of the coverage purchased while receiving protection priced on the full value of the policy. A subsidy of that proportion changed which coverage levels were worth buying and, at the extensive margin, whether cropping a risky field promised a positive expected return at all. The Agricultural Risk Protection Act of 2000 raised subsidy rates, and participation expanded until it reached roughly 90 percent of eligible acres. At that breadth, the insurance title became part of the standard financial architecture within which planting decisions across most eligible land were made, and an instrument used on nearly all eligible acres did not need a large effect on any single acre to produce a consequential effect in aggregate.

Roger Claassen, Fernando Carriazo and Joseph C. Cooper, in research published by the Economic Research Service of the United States Department of Agriculture in 2011, estimated that premium subsidies shifted land into crop production. Their national estimate was modest, and the modesty deserves the same careful reading given to the decoupled payment literature. A modest national average concealed a concentrated geography. The shift into cropping occurred disproportionately where insurance made cropping viable on marginal land, meaning land whose soil, moisture or climate left its return to cropping near the threshold between profit and loss under uninsured conditions. On highly productive land, cropping was profitable with or without subsidized insurance, and the subsidy largely transferred income on acres whose use did not change. On land far below the threshold, even subsidized insurance could not make cropping pay. Between those zones lay the acres where the subsidy was decisive, and on those acres the production effect and the land use effect were the same event.

Simulation work focused on the Northern Plains, associated with Claassen and coauthors, exposed the price dependence of that conversion effect. Insurance subsidies materially raised grassland conversion under low price scenarios and mattered little under high price scenarios. The asymmetry followed from the threshold logic. When crop prices stood high, expected returns pulled marginal grassland toward cropping on market terms alone, and the subsidy added protection to a conversion that prices had already motivated. When crop prices stood low, market returns alone left grassland in grass, and the subsidy became the factor that could lift the insured return to cropping above the return to leaving the land undisturbed. A study run during strong prices would therefore record small conversion effects, while the same subsidy structure held a larger effect in reserve for weak price seasons.

Specific features of coverage deepened the effect on the riskiest land. Prevented planting coverage paid when weather conditions kept a producer from planting an insured crop, and revenue and yield guarantees cushioned the outcome when a planted crop performed badly. Together those features raised the expected return of cropping drought prone and wet land relative to the return of grass. Land that flooded in wet springs or failed in dry summers imposed losses on an uninsured producer with a frequency that made grass the rational use over a run of seasons. Insurance truncated the worst of those outcomes and compensated some seasons in which planting never occurred, so the mean return rose, the lower tail thinned, and a field whose uninsured history argued for permanent grass could present an insured profile that argued for the plow.

That aggregate was where the production finding became a conservation finding, and the connection was direct rather than rhetorical. Land converted from grass to crops under the influence of insurance subsidies was frequently land whose environmental value in grass had been the reason conservation policy valued it. Grassland held soil in place, stored carbon in roots and soil, filtered water moving toward streams and provided habitat that cropped fields did not support. A subsidy structure that raised conversion pressure on exactly the marginal acres where grass competed closely with crops therefore worked against the land retirement and grassland objectives pursued elsewhere in the same farm bills. The commodity and insurance titles pushed at one margin while the conservation title paid at another, and the production literature and the conservation literature described the same acres from opposite sides.

Concentration at the margin was the recurring signature across both instruments. Decoupled payments moved production a little, mostly through the finances of constrained and risk averse operations. Insurance subsidies moved land use a little in national terms and a great deal on the particular acres where the decision was close. In both cases the average effect understated the effect at the margin that mattered, and a debate conducted only in national averages missed where the programs actually changed the landscape. The reform of 1996 had succeeded in reducing the production distortion carried by each dollar of commodity support, and the studies by Hennessy, by Goodwin and Mishra, by Weber and Key and by Serra, Goodwin and Featherstone together explained why a residue of distortion survived and why that residue stayed small. The growth of the insurance title had meanwhile rebuilt a production influence through a different door, attaching support to present season decisions on present season acres and altering the fate of marginal grassland in the process. Farm subsidy law did change what was planted and what was produced, least where payments were fixed on history and most consequentially where a close decision about fragile land waited for a reason to break toward the crop.

Finding Four: A Cotton Dispute That Priced Domestic Support

The cotton dispute began as a claim about a single commodity, and it ended as the most complete external accounting farm subsidy law received during the period covered by the record. Brazil, a competing cotton exporter, requested consultations at the World Trade Organization in September 2002 against United States cotton support, and the proceeding that followed was given the designation DS267. The request challenged a program structure rather than a single payment, and the distinction shaped everything that followed. Brazil did not ask a tribunal to decide whether farm support existed in the abstract. Brazil asked whether specific instruments, operating in specific ways, crossed specific lines that trade law drew between support a government could provide freely, support a government could provide only if the support caused no adverse effects to another member, and support a government had agreed not to provide at all. The case therefore forced a kind of precision that domestic debate rarely required. In domestic debate, support could be defended or opposed as a whole, praised as a safety net or condemned as waste, without separating one payment mechanism from another. In Geneva, separation was the entire exercise, and the postwar trade legislation sequence that built the system inside which the dispute was heard supplied the institutional setting in which that separation had legal consequences.

Cotton served as the test case for reasons that were structural rather than symbolic. Cotton was a traded commodity, produced in the United States for a market that extended far beyond domestic mills, priced through transactions that connected growers in different countries to the same buyers, and exported in volumes large enough for changes in United States production and sales to register in world prices. A complainant challenging support for a commodity consumed almost entirely at home would have faced a different and harder evidentiary problem, because the chain from a domestic payment to harm in another country would have run through indirect channels that were difficult to isolate. Cotton shortened that chain. Brazil produced cotton for the same world market, exported into that market, and could present its claim as a claim about price formation in a market both countries shared. If payments in the United States kept production higher, or kept exports flowing at volumes or prices that would not have prevailed without the payments, the effect would appear where Brazil sold its own crop. Serious prejudice, the legal concept at the center of the actionable subsidy findings, required exactly that kind of showing, and cotton gave Brazil a factual setting in which the showing could be attempted with production data, export data, and price evidence drawn from one identifiable market.

The panel examined each instrument separately, Step 2 payments, GSM 102 export credit guarantees, marketing loan and counter cyclical payments, and direct payments, and the findings in the panel report of September 2004 and the Appellate Body report of March 2005 preserved those distinctions. Collapsing the findings into a single sentence about illegality would erase that work, because the reports placed the instruments in different legal categories and attached different consequences to each category.

What did the cotton dispute actually decide about United States farm subsidies?

The panel and Appellate Body held in 2004 and 2005 that Step 2 payments and GSM 102 guarantees were prohibited subsidies, that marketing loan and counter cyclical payments were actionable subsidies causing serious prejudice through world price suppression, and that direct payments failed green box criteria in part.

The prohibited category occupied the strictest position in the framework the reports applied. A prohibited subsidy was not a subsidy that a member could maintain upon a showing that the subsidy caused no harm. The prohibition attached to the design itself, to payments contingent on export performance and to payments contingent on the use of domestic rather than imported goods, and the remedy logic followed from that attachment. Findings in the prohibited category identified instruments to be withdrawn, not instruments to be recalibrated until the effects became acceptable. Step 2 payments fell into that category on two grounds stated in the findings. Payments made in connection with cotton exports operated as export subsidies, because the payment structure rewarded the export transaction itself. Payments made to domestic users of cotton operated as import substitution payments, because the payment structure rewarded the choice of domestic cotton over imported cotton. The dual character of the Step 2 mechanism, reaching exporters on one side and domestic users on the other, placed the mechanism on both prohibited paths at once, and the findings recorded both paths rather than selecting one as sufficient.

The GSM 102 export credit guarantees received a prohibited finding through a different route that reached a parallel result. An export credit guarantee did not resemble a direct cash payment, and that difference in form could have supported an argument that the guarantee belonged in a gentler category. The findings looked through form to function. As the reports characterized the program, the guarantees operated as prohibited export subsidies, because the terms on which the guarantees were provided conferred a benefit tied to exports that the governing disciplines did not permit. A government could deliver export support through credit mechanisms and risk assumption rather than through appropriated payments alone, and the reports examined what the guarantee did for the export transaction. The categorization followed the operation rather than the label.

Marketing loan payments and counter cyclical payments occupied a different position, and precision about that position mattered as much as precision about the prohibited findings. The reports treated those payments as actionable subsidies, a category that did not forbid the payments by design. Actionable subsidies became vulnerable when a complaining member demonstrated adverse effects, and Brazil demonstrated an effect of a specific kind. The findings held that the payments caused serious prejudice to Brazil through significant suppression of world cotton prices. Every element of that sentence carried weight. The prejudice was serious, a threshold concept rather than a synonym for any measurable influence. The mechanism was price suppression, a finding that world prices were lower than prices would have been without the payments, rather than a finding that prices fell in absolute terms during a particular period. The scope was the world market for cotton, the market in which Brazilian growers and exporters sold. A payment could therefore survive scrutiny as an actionable subsidy in one market context and fail in another, because the category turned on demonstrated effects. The marketing loan and counter cyclical findings turned on the price linkage built into both instruments, payments that rose as market prices fell and that thereby sustained production incentives at moments when price signals alone would have contracted output. The design that domestic law valued as counter cyclical responsiveness became, in the trade analysis, the design that connected support to the suppression of the price itself.

Direct payments produced a third kind of finding, narrower than the prohibited determinations and different in character from the serious prejudice determination. Direct payments had been constructed as payments based on historical production, and the construction supported a claim that the payments belonged in the green box, the category for support considered minimally trade distorting because the support was decoupled from production. The reports rejected the green box claim in part, and the ground stated in the findings was concrete. Producers who received direct payments were barred from planting fruits and vegetables on base acres. A payment described as decoupled nevertheless carried a condition that governed a planting decision, restricting one use of the land to which the payment attached. The restriction reached only a defined set of crops, and the findings treated the restriction with corresponding care, rejecting the green box classification in part rather than declaring the payments equivalent to price linked support. The lesson embedded in the direct payments finding concerned the fragility of decoupling as a legal status. Decoupling was not established by the payment formula alone. Conditions attached to the payment counted, and a single planting restriction, maintained for domestic program reasons, was enough to compromise a classification that the payment formula appeared to secure.

Read together, the three kinds of finding composed a map rather than a single verdict. A prohibited finding pointed toward repeal or fundamental restructuring. An actionable finding pointed toward removal of the adverse effects, leaving more room for argument about whether changed conditions had cured the prejudice. A green box failure pointed toward adjustment of conditions attached to a payment. The reports supplied a classification of cotton support by legal vulnerability that domestic legislation then had to answer, instrument by instrument.

The first legislative answer arrived through repeal of the instrument carrying the clearest prohibited finding. The United States repealed Step 2 effective August 1, 2006, in the Deficit Reduction Act of 2005. The sequence deserves attention for what repeal did and did not accomplish. Repeal removed the payment mechanism whose export and import substitution features had produced the prohibited determinations, and removal of a repealed mechanism was the most complete form of compliance available for that instrument, because no redesigned version remained to be tested against the prohibition. Repeal did not, by itself, resolve the actionable subsidy findings, because marketing loan and counter cyclical payments rested on a different legal basis and continued in forms that Brazil regarded as preserving the prejudice the reports had identified. Repeal also did not resolve the GSM 102 question in the manner Brazil sought, because the guarantees operated through program terms that could be adjusted by degree, and adjustment by degree invited disagreement about sufficiency. The compliance proceedings that followed grew out of exactly that structure. One instrument had been eliminated outright, while other instruments persisted in modified or unmodified forms whose effects remained in dispute, and the legal framework provided a procedure for testing whether the measures taken to comply had in fact brought the member into conformity.

Compliance proceedings in 2007 and 2008 produced findings that the United States had not fully complied and that serious prejudice continued. The findings demonstrated that the serious prejudice determination was not treated as a historical judgment confined to the marketing years examined in the original proceeding. Cotton was planted and sold in successive seasons, world prices formed in successive seasons, and a support structure that remained in place generated the alleged prejudice anew with each crop. Compliance review captured that recurrence and gave the recurrence legal expression.

Failure to achieve a compliance finding satisfactory to Brazil opened the path to countermeasures, and the arbitration of 2009 converted the dispute from a contest over characterization into a contest over an authorized amount. Arbitration authorized Brazilian countermeasures with a fixed component of 147.3 million dollars per year plus a further variable amount. Authorization did not itself impose the countermeasures. Authorization established that Brazil could suspend concessions up to the determined level without itself breaching the governing obligations, shifting the practical burden of the dispute. After authorization, Brazil held a priced entitlement to respond, and the United States faced a choice among redesign sufficient to end the entitlement, negotiation sufficient to suspend the exercise of the entitlement, or acceptance of countermeasures applied against United States trade interests that had no connection to cotton production.

Brazil prepared retaliation covering roughly 830 million dollars of trade in goods and services, and the preparation included measures directed against intellectual property. The breadth of the prepared response altered the domestic politics of the dispute. Countermeasures against goods trade widened the circle of affected interests to exporters of unrelated products whose market access in Brazil stood at risk because of a cotton program those exporters did not design and did not receive. Countermeasures reaching intellectual property widened the circle further, toward holders of rights whose protection in Brazil could be suspended as a lawful response to agricultural support. Cotton growers continued to receive support under domestic law, but the authorized response placed identifiable costs on unrelated domestic interests, and that imbalance became part of the environment in which settlement was negotiated.

The settlement path began with the Framework Agreement of June 2010, under which Brazil held retaliation in abeyance and the United States made annual payments directed, remarkably, to the cotton sector of the complaining country. The United States paid 147.3 million dollars per year to the Brazilian Cotton Institute, the Instituto Brasileiro do Algodão, for technical assistance and capacity building in Brazil’s cotton sector. The figure matched the fixed component of the arbitration authorization, and the correspondence was not coincidental in structure. A sum that arbitration had established as an annual measure of authorized countermeasures became, by agreement, an annual transfer, and the transfer purchased suspension of the retaliation the sum had measured. The destination of the payments gave the arrangement its distinctive character. Compensation in trade disputes ordinarily takes forms such as expanded market access or payments into general funds. The framework directed funds to an institute serving the very industry that had brought the complaint, financing technical assistance and capacity building for Brazilian cotton, the sector whose competitive position the original findings had found prejudiced by United States support. The United States thereby funded improvement of a competitor’s productive capacity as the interim price of maintaining domestic programs that the settlement had not yet redesigned to Brazil’s satisfaction.

Payments under the framework totaled 505.5 million dollars before payments stopped in October 2013 amid the federal budget sequester and shutdown. The cessation revealed a vulnerability built into the arrangement from the beginning. An international understanding that depended on annual payments depended equally on the domestic budget processes that produced those payments each year, and a budget conflict unrelated to cotton interrupted a commitment made to resolve a cotton dispute. Brazil had suspended the exercise of authorized countermeasures in reliance on a payment stream, and the payment stream proved subject to fiscal events in the paying country that the receiving country could neither predict nor control. Any successor arrangement had to answer that fragility.

The Memorandum of Understanding of October 1, 2014 answered the fragility of annual payments with finality of a defined kind. The United States made a one time final payment of 300 million dollars to the Brazil Cotton Institute. Brazil relinquished countermeasure rights arising from the dispute and accepted a peace clause covering United States cotton support for the life of the 2014 farm bill, Public Law 113-79, provided programs operated within the agreed terms. GSM 102 fees and tenor were reformed as part of the settlement architecture, addressing the export credit instrument that had carried a prohibited finding alongside the Step 2 payments. Total United States payments to Brazil under the settlement path reached 805.5 million dollars, a figure composed of the 505.5 million dollars paid under the framework and the 300 million dollars paid under the memorandum. The structure of the 2014 resolution differed from the structure of the 2010 framework in each element that had proven unstable. A single payment replaced an annual stream and removed the settlement from dependence on successive budget cycles. Relinquishment of countermeasure rights replaced suspension of retaliation and closed the remedy Brazil had held since the 2009 arbitration. A peace clause bounded by the life of a named statute replaced an open ended interim period, tying the legal truce to the legislation that defined the support programs in operation during that period.

The consequence of the sequence can be stated plainly and neutrally, without embellishment in either direction. A domestic support design produced a documented, priced international consequence. The documentation ran through a panel report, an Appellate Body report, compliance findings, and an arbitration award, each stage narrowing the range of plausible denial about what the findings were. The price ran through an authorized countermeasure figure, annual payments totaling 505.5 million dollars, and a final payment of 300 million dollars, for a combined 805.5 million dollars transferred to the complaining country’s cotton institute. The United States paid that sum as the price of keeping its own program substantially in place during the years the settlement covered, because the framework and the memorandum did not require dismantlement of commodity support for cotton as a condition of peace. The payments resolved a dispute about support by compensating the party the support had prejudiced, while the underlying question of how cotton support should be designed received its answer through a separate process, the ordinary legislative process of farm bill revision, operating on a timetable the settlement deliberately accommodated through the peace clause.

Program changes arrived with the Agricultural Act of 2014, and the changes reflected the dispute in design even where the statute answered broader pressures as well. The 2014 act ended direct payments and counter cyclical payments generally, removing across commodities the two instruments that had carried, respectively, the green box failure and part of the serious prejudice finding in the cotton case. Upland cotton was excluded from Agriculture Risk Coverage and Price Loss Coverage, the new commodity title programs the act created, a carve out without parallel among the major program crops and one that tracked the litigation history of cotton specifically. In place of commodity title coverage, Congress created the Stacked Income Protection Plan, known as STAX, a county level revenue insurance product for cotton carrying an 80 percent premium subsidy. Cotton support after the 2014 act therefore rested principally in the insurance title rather than the commodity title, delivered as premium support for a revenue product triggered at the county level rather than as payments calculated under the commodity programs available to other covered commodities. The Bipartisan Budget Act of 2018 then made seed cotton a covered commodity eligible for Agriculture Risk Coverage and Price Loss Coverage beginning with the 2018 crop, restoring a commodity title pathway for cotton in a redefined form, with eligibility attached to seed cotton as the covered commodity rather than to upland cotton under the earlier structure.

The design lesson the sequence teaches emerges from the direction of migration across those changes, stated as analysis grounded in the sequence rather than as judgment on the merits of any instrument. Support for cotton moved, under legal pressure concentrated on the commodity title, toward the insurance title, and the insurance title operated under disciplines that were less tested at the vantage of January 15, 2021. The commodity title instruments had been examined element by element in DS267, their payment triggers, planting conditions, and price effects classified and, in part, condemned. No comparable examination of a premium subsidized revenue insurance product of the STAX type stood in the record at that vantage. Migration toward a less tested legal environment was a rational response to the findings as far as the findings reached, and the migration simultaneously deferred the hardest questions. A premium subsidy that enabled a grower to purchase revenue protection at a fraction of the actuarial cost influenced production decisions through risk reduction rather than through a payment formula, and the channels of influence, wealth effects, insurance effects, and the planting of marginal land that the production literature examined, operated whether or not a tribunal had classified the instrument. The 2018 restoration of commodity title eligibility through seed cotton added a further layer, combining an insurance centered structure with renewed access to the programs from which upland cotton had been excluded in 2014, and the combined structure was the arrangement in place when the peace clause tied to the 2014 act approached the end of the statutory life that defined the clause.

Finding Four therefore stands in the record as the point at which farm subsidy law acquired an external price tag with a payee, a schedule, and a total. The total of 805.5 million dollars did not measure the full effect of the programs on world cotton markets, and the arbitration figures were never presented as such a measure. The total measured the negotiated and adjudicated price of non compliance and settlement across a defined period, which is a narrower and more durable kind of fact. Programs designed in Washington were examined in Geneva, priced in arbitration, paid through an institute in Brazil, redesigned in the 2014 act, and partially restored in 2018, and each stage of that sequence is documented in instruments that remain available for inspection. That documentation, rather than any characterization imposed upon the sequence, is the finding.

Finding Five: Conservation Outcomes Measured Against the Land Retired

The conservation title produced the clearest physical record of any part of the farm support system because its central land retirement program changed a visible condition on identifiable acres. A field taken out of crop production and placed under protective cover either eroded or did not erode, either discharged sediment and nutrients toward a waterway or retained a larger share of both, and either offered cover to wildlife or returned to tillage and harvest that offered less of it. Farm Service Agency reporting credited land retirement with large reductions in soil loss and nutrient movement, alongside restored wetlands, expanded riparian buffers, greenhouse gas sequestration and improved habitat. The same record carried qualifications that defined the finding as firmly as the benefits did, since enrollment depended on rental rates that had to compete with crop returns, protection depended on contracts of limited term, and the working lands programs rested on a harder causal question that Economic Research Service research left open at the January 2021 vantage.

The Conservation Reserve Program was created by the Food Security Act of 1985 and was administered by the Farm Service Agency. The federal government offered annual rental payments under contracts running 10 to 15 years, and the producer retired environmentally sensitive cropland from crop production for the contract term. Enrolled land was not simply idled. Retirement meant a shift from annual cropping to a protective cover intended to hold soil in place, slow water moving across the surface, filter sediment and nutrients before they reached streams, and create a more stable habitat structure than a field planted and harvested in a yearly cycle. The rental payment compensated the producer for taking land out of production, while the cover established on that land supplied the public environmental return.

While a contract remained in force and the cover remained intact, the environmental mechanism operated continuously rather than through a single project or a one time change in technique. Roots held soil through seasons when a cropped field might have stood bare between harvest and planting. Vegetation slowed runoff and gave sediment a place to settle before it entered a ditch or stream. Undisturbed cover gave ground nesting birds and other wildlife stability across successive breeding seasons. The benefits accumulated because the land use itself had changed for the length of the contract, and that accumulation formed the basis for the large cumulative totals in Farm Service Agency reporting.

Does the Conservation Reserve Program actually reduce soil erosion?

Farm Service Agency reporting credited the Conservation Reserve Program with preventing more than 9 billion tons of soil erosion since its creation, reducing erosion by roughly 325 million tons per year relative to pre-program levels on enrolled land retired from cropping under long term contracts.

The soil figures carried weight because erosion was the outcome most closely tied to the act of retirement itself. Cropland identified as environmentally sensitive tended to be land where slope, soil texture, or exposure to wind or water made repeated cultivation costly in soil loss. Removing tillage and maintaining cover addressed the exposure directly. Soil that stayed on the field preserved the productive base that erosion would otherwise have carried away, and soil that stayed on the field also stayed out of drainage channels, streams, rivers and lakes, where sediment degraded water quality and aquatic habitat. The comparison framed the benefit as the distance between enrolled land under retirement cover and the same land under its earlier cropped condition.

The cumulative figure reflected duration as much as intensity on any single acre. Contracts running 10 to 15 years allowed cover to establish and persist through weather cycles that tested it in different ways. A cumulative total described what retirement had prevented across the years of participation, and it did not guarantee that a given acre would retain the same protection after a contract ended, when the future erosion path depended on the management chosen after expiration.

Water quality formed the second major strand of the documented record, and it followed from the soil record through both sediment and dissolved nutrients. Farm Service Agency reporting stated that the program kept more than 600 million pounds of nitrogen and more than 100 million pounds of phosphorus out of streams, rivers and lakes each year. Phosphorus often travelled attached to sediment, so holding soil in place reduced phosphorus movement at the same time. Nitrogen moved more readily with water, so slower movement of water through established cover and buffer vegetation, together with plant uptake on land no longer managed under a cropped field pattern, reduced the load reaching waterways. The two figures described connected effects of the same land use change.

Wetlands and riparian buffers extended that connection from the field interior to the places where fields met water. Farm Service Agency reporting credited the program with restoring more than 2 million acres of wetlands and more than 2 million acres of riparian buffers. A restored wetland held water on the landscape, trapped sediment, processed nutrients and supplied habitat of a kind that drained or cropped wetland ground could not supply. A riparian buffer placed vegetation between cropped land and a stream or river, intercepting runoff before it reached open water and stabilizing banks where erosion at the channel edge would otherwise have added to the sediment load. Those acres mattered beyond their share of total enrollment because location governed environmental return, and an acre positioned to intercept runoff toward a waterway delivered services that an acre far from water could not deliver in the same measure.

Climate accounting added a further reported benefit, expressed as sequestration rather than avoided loss. United States Department of Agriculture releases credited the program with sequestering an annual average of about 49 million tons of greenhouse gases. Land taken out of annual cultivation and placed under persistent cover stored carbon in plant material and in soil, and it avoided some emissions associated with repeated tillage and cropping on the retired acres. The figure was an annual average in the department accounting, and that averaging mattered for honest reading, because sequestration on a given acre varied with the cover established, the soil condition at enrollment and the climate of the region. The accounting compared retired land with a counterfactual in which the land remained in crop production, so the reported gain represented the difference between those two states rather than an absolute property of grass cover alone.

Wildlife habitat completed the set of benefits that Farm Service Agency reporting and Congressional Research Service summaries treated as part of the documented record. Pheasants, quail, ducks and grassland birds used the cover, wetland and buffer habitats that retirement created, and the value of that habitat rested on the same persistence that produced the soil and water gains. A field retired for 10 to 15 years offered nesting cover, brood rearing cover, winter cover and feeding areas across repeated seasons, rather than the brief intervals between operations in a cropped field. Wetland restoration supplied waterfowl habitat directly, and grass cover on upland acres supplied the nesting and sheltering structure that grassland birds required and that intensive annual cropping compressed into field edges and fragments.

The rental payments that purchased those outcomes ran about 1.8 billion dollars annually, according to the Farm Service Agency figure carried in the record. That expenditure formed the cost side of the clearest benefit accounting in the conservation title, because the payment and the land use change arrived together under a contract that specified both. The government did not pay for a promise of effort or for a practice whose adoption might have occurred on the same timetable without support. Payment followed enrollment of eligible land and the obligation to keep that land out of crop production under approved cover for the contract term, so compliance was observable in the condition of the land itself. That observability gave the program a firmer evidentiary footing than payment streams whose environmental action was indirect, even though rental rates, targeting and duration all influenced the return obtained for the expenditure.

The measurement posture deserved explicit statement because it explained why land retirement benefits were the most directly documented in the conservation title. Agency reported physical outcomes were measured against the counterfactual of the same land cropped. Erosion prevented meant erosion estimated not to have occurred because enrolled acres carried retirement cover instead of a crop. Nutrients kept out of waterways meant loads estimated to be lower than the loads the same acres would have generated under cropping. Greenhouse gases sequestered meant storage and avoided emissions estimated against a cropped baseline. Counterfactual measurement was not a defect unique to the program, since any claim about prevention required a comparison with a state of the world that did not occur during enrollment, but the comparison was unusually close to the intervention. The treatment and the outcome occupied the same acres and the same years, the land use change was total rather than marginal, and the physical processes linking cover to reduced erosion and runoff were direct even where national totals required estimation.

Scale set the boundary within which those benefits operated at the January 2021 vantage. Enrollment stood at roughly 22 million acres in 2020, against a statutory cap that the Agriculture Improvement Act of 2018 phased from 24 million acres toward 27 million acres by 2023. The gap between enrollment and cap directed attention to the economics of participation, because a cap authorized acres and did not enroll them. Producers enrolled land when the rental payment outweighed the return expected from cropping, after accounting for production costs, yield prospects, price expectations and the flexibility surrendered for a contract term of 10 to 15 years. The rental offer competed against a modest expected crop return on marginally productive land and against a far stronger return on productive land.

Rental competition therefore made enrollment sensitive to commodity prices in a way that ran against the timing of environmental need. When commodity prices rose, crop returns rose with them, and producers had a stronger reason to keep acres in production or to return expiring acres to production rather than enroll or reenroll them at rental rates formed under earlier conditions. Enrollment fell under that pressure unless rental rates adjusted enough to restore the balance, and rental adjustment itself faced budget and program design limits. Fragile land remained fragile in a high price year, and it was precisely in a high price year that the incentive to crop it intensified.

Permanence formed the standing qualification on every benefit figure in the record. Benefits lapsed when contracts expired and land returned to production. Grass cover plowed under at the end of a contract released part of the protection that enrollment had built, soil again stood exposed to the erosion processes that retirement had interrupted, nutrient and sediment movement resumed along the paths that cropping reopened, and habitat that had depended on undisturbed cover across successive seasons contracted to the margins that production allowed. None of that cancelled the gains recorded during the enrollment years, since erosion prevented in a given year and nutrients kept out of a waterway in that year were outcomes that had occurred, but expiration separated the achievement from a permanent transformation of the land. Protection was rented for a term, and the environmental return lasted in full only as long as the rental relationship or a successor protective use endured. The cumulative total likewise represented accumulation across contracts that began and ended at different times on different acres, not the creation of a single protected estate held intact at the end of the period.

Working Lands and the Additionality Question

The Environmental Quality Incentives Program, dating to 1996, and the Conservation Stewardship Program, authorized in 2002 and renamed in 2008, paid for conservation practices on land that remained in production. The payment supported a change in how an acre was farmed rather than a decision to stop farming it for a contract term. Working lands programs occupied the far larger agricultural landscape that land retirement did not touch, and their design answered a real limitation of a retirement only approach, because most farmland was going to remain in production under any plausible program structure at the January 2021 vantage. Cost sharing and stewardship payments lowered the producer cost of adopting practices intended to reduce nutrient loss, improve soil condition, manage water or protect habitat on working acres, attempting to move a practice from unadopted to adopted, or from a lower level of conservation performance to a higher one, while the acre continued to produce a crop or support livestock. Retirement addressed the most sensitive acres by removing them from production, and working lands payments addressed the environmental performance of the acres that production retained.

The causal structure, however, differed in a way that changed the evidentiary standing of the results. Under land retirement, the counterfactual comparison remained close to observation because the enrolled acre was visibly out of production. Under a working lands payment, the relevant comparison asked what the same producer would have done on the same acre, in the same period, without the payment. A practice recorded as cost shared might have been adopted later without support, adopted in a smaller or different form, or adopted at the same time and scale because the producer had already decided that the practice paid for itself. Payment records established that a practice had been installed or adopted with public support, and they could not by themselves establish that the support had caused the adoption. That gap between a documented practice and a caused practice was the additionality problem. Additionality asked whether the payment caused a practice that would not otherwise have occurred. Where a payment was fully additional, the expenditure purchased a conservation gain that the counterfactual without payment would not have produced. Where a payment supported a practice that the producer would have adopted anyway on the same schedule, the expenditure transferred income but did not purchase that environmental gain, because the gain belonged to the counterfactual as well as to the observed outcome. Between those poles lay partial additionality, where payment accelerated adoption, enlarged its scale, or raised its quality or durability, each form requiring evidence beyond a count of practices paid for.

Economic Research Service research gave that problem empirical content rather than leaving it as a theoretical caution. Research by Roger Claassen and coauthors found evidence that a share of cost shared practices would likely have been adopted without the payment. That finding did not state that all or most supported practices lacked additionality, and the record framed it as evidence about a share, with the size of that share varying across practices and settings rather than settling into a single rate applicable to working lands support as a whole. A practice with strong private returns stood a greater chance of adoption without support than a practice whose benefits accrued mainly off the farm in cleaner water or habitat that the producer could not capture as income. The Claassen research therefore disciplined benefit accounting in both directions, blocking any assumption that every paid practice represented a program caused gain and equally blocking any claim that payments bought nothing because some practices would have occurred anyway.

The same research posture held that benefit estimates for working lands programs were less firmly established than for land retirement, a comparative judgment that followed from the structure of the two approaches. Land retirement changed land use in a discrete and observable way, so the physical chain from intervention to outcome ran over a short distance. Working lands support changed practices within continuing production systems where weather, prices, technology, producer learning and other programs also moved behavior at the same time, so isolating the payment effect required separating one influence among several. At the January 2021 vantage, additionality remained a genuinely open research question, and the record supported no verdict in either direction beyond the documented finding that a share of cost shared practices would likely have been adopted without payment. Working lands payments offered reach across the producing landscape and the possibility of durable management change outlasting a payment period, but the record had not established how common that pathway was.

Conservation compliance supplied the connection that tied the conservation title back to the commodity and crop insurance titles, and it operated through eligibility rather than through a payment for a practice or a rental contract. Conditions known as sodbuster and swampbuster had been attached to program eligibility since 1985. In qualitative terms, sodbuster addressed the bringing of highly erodible land into production without an approved conservation system, and swampbuster addressed the conversion of wetlands for crop production, with loss of eligibility for farm program benefits as the sanction attached to covered violations. Compliance therefore used access to the support system itself as leverage over land use decisions that no conservation payment had purchased.

The full system, read at the January 2021 vantage, therefore contained currents running in opposite directions inside the same statute. Finding Three identified the tension in measured form on the production side, where insurance support raised the expected return of cropping risky and marginal land and thereby pulled such land toward production. The conservation title paid in the other direction, retiring environmentally sensitive cropland under the Conservation Reserve Program and paying for protective practices on working land under the Environmental Quality Incentives Program and the Conservation Stewardship Program, while compliance conditions penalized specified conversions through eligibility. Premium support improved the economics of planting a marginal acre, a rental offer improved the economics of retiring an eligible acre, and compliance altered the economics of converting protected categories of land in the first place. Which force prevailed on a given acre depended on prices, yields, rental rates, insurance terms and the physical character of the land. The conservation achievement described in Farm Service Agency reporting was therefore substantial and bounded at once, a balance struck acre by acre and contract by contract rather than a direction imposed on American farmland as a whole.

The Five-Finding Evidence Table

Finding	Direction of effect	Period measured	Principal research	Mechanism	Rating
Concentration of payments	Payments concentrated at the top	1995 to 2016	Environmental Working Group database of USDA records	Payments keyed to acres, yields and production scale with size	Settled
Capitalization into land values and rents	Rents and land values raised	2003 to 2009 studies	Kirwan (2009); Goodwin, Mishra and Ortalo-Magné (2003)	Payment streams priced into rents and sale values	Settled
Production effects of decoupled payments and insurance subsidies	Small for decoupled payments, larger for insurance	1998 to 2012 studies	Goodwin and Mishra (2006); Claassen and coauthors (2011)	Wealth, credit and insurance channels alter marginal planting returns	Settled
International consequences (the cotton dispute)	Serious prejudice found, settlement paid	2002 to 2014	WTO panel and Appellate Body reports, DS267	Price suppression from marketing loan and counter-cyclical payments	Settled
Conservation outcomes and additionality	Retired land benefits documented, working lands additionality open	1985 to 2020	Farm Service Agency reporting; Claassen and coauthors (ERS)	Rental contracts retire fragile land; cost share may fund adopted practices	Contested

The Case the Evidence Has to Answer

The strongest case for the programs begins with the structure of agricultural risk, and it deserves to be stated at full strength before any finding in this article is allowed to count against it. Farming joins yield risk and price risk in a combination that private insurance markets handle badly. A drought does not strike one policyholder in a county while sparing the neighbor whose premiums balance the loss. Weather, pests and disease move across regions, so claims arrive together in the worst years, and prices often move with the same harvest conditions that determine yields, compounding a revenue loss or masking it in ways that differ by crop and by year. Correlated risk of that kind is the precise condition under which private insurance struggles, because the insurer cannot diversify the exposure away across independent policyholders and must either hold large reserves, charge premiums that reflect the correlated tail, or withdraw. Adverse selection compounds the failure. The producer who knows that a field floods in wet years or burns in dry ones has the strongest reason to buy coverage and the clearest information advantage over the insurer pricing it, so a voluntary unsubsidized market tends to draw the riskiest acres at prices the safer acres will not pay, and coverage thins toward the operations that need it least. A public role in agricultural insurance is not, on that account, an anomaly in need of apology. It is a response to two named market failures that the private market, left alone, had demonstrated across decades of thin participation and abandoned products.

Income volatility supplies the second strand of the case, and its scale is easy to understate in a table of annual averages. Net farm income swings by tens of billions of dollars between adjacent years, driven by export demand, energy prices, exchange rates and weather on several continents at once, and an individual operation experiences that national swing amplified by its own yields and its own marketing decisions. A business with heavy fixed costs in land, machinery and debt service cannot treat a revenue collapse as a signal to contract gracefully for a season. The machinery loan, the operating line and the rent fall due in the bad year on the same schedule as in the good one. Lenders understand the arithmetic and lend against expected support, so the presence of a predictable support framework lowers the cost and raises the availability of credit across the sector, including for borrowers whose own payments are modest. Remove the framework abruptly and the effect would not be confined to payment recipients. Credit terms would tighten against the expected income that had collateralized them, and operations carrying debt priced on the old expectations would face refinancing into a weaker income outlook at the moment of greatest exposure.

The historical comparison strengthens the point, because the alternative to standing programs was tried for decades and its record is part of the evidence. Before the modern crop insurance framework matured, the federal response to widespread crop failure ran through ad hoc disaster legislation, assembled after the loss, debated while operations waited, and paid on terms that varied with the politics of the particular disaster and the composition of the Congress that answered it. That system was slower than a standing program, since no payment could move until a bill passed. It was less predictable, since producers could not know in advance whether a given loss in a given year would attract a bill at all, or on what conditions. And it was no cheaper, because a legislature confronting photographed losses across several states faces pressure to pay generously and broadly, without the premium contributions, coverage conditions and loss adjustment discipline that an insurance structure imposes before the event. A standing program with published rules converts an unpredictable political transfer into a plannable business input, and that conversion has value to producers, lenders and rural communities apart from the size of any single payment.

The capacity argument carries the case beyond the farm balance sheet. Price troughs in agriculture are often long enough to force exits that cannot be reversed when prices recover. An operation that sells its land, disperses its breeding herd, loses its skilled labor and surrenders its place in a local processing or marketing chain does not reassemble when the market turns. Land converted out of agriculture rarely returns, specialized equipment sells at distress prices into a falling market, and the knowledge held by an experienced operator leaves the sector with the operator. Support that carries a producer base through a trough therefore preserves productive capacity as a standing national asset, and food security framed that way is not a slogan but a claim about irreversibility. A nation that allows its production base to liquidate in every deep trough buys its recovery later, at higher cost, with greater import dependence in the interval. The statutes were written against memories of exactly that liquidation, and the stabilization aim they state reflects a judgment that the social cost of lost capacity exceeds the fiscal cost of bridging income through the years that would destroy it.

None of that case, however, is a case about grocery prices, and the food price passage in the evidence cuts against advocates and opponents with equal force. The United States Department of Agriculture Economic Research Service Food Dollar Series put the farm share of the consumer food dollar at 14.6 cents in 2017, the lowest figure since the series began in 1993, with roughly 85 cents of each food dollar going to processing, transport, wholesaling, retailing and food service. The arithmetic that follows is unforgiving in both directions. If the farm gate receives about fifteen cents of the retail dollar, then even a large change in the support attached to farm production moves through a small base before it reaches the shelf, diluted further by the processing and distribution costs that make up the rest of the price. Claims that subsidies make groceries cheap therefore fail, because the supported commodities enter retail food prices through too narrow a channel to account for the affordability of the American food supply, which rests far more on productivity, logistics and competition in processing and retail. Claims that ending subsidies would make groceries expensive fail on the same arithmetic, since removing support from a fifteen cent share cannot produce a large movement in a dollar dominated by other costs. The standing exception class proves the rule by its design rather than by the general run of programs: a small class of programs works by raising domestic prices, with the sugar program, built on import restrictions and price supports, as the described example of support delivered through the price mechanism instead of through the Treasury. That exception matters analytically because it shows what a program would have to do to move consumer prices at all, namely restrict supply and hold a domestic price above the world level, which is the opposite of the income payment structure that dominates the commodity and insurance titles.

The evidence also has to answer a fairness point about who the critique in the research literature actually targets. Read closely, the strongest critique in that literature is aimed at design and distribution rather than at the existence of risk management support. Its counts are specific. Payments scale with size because they are keyed to acres, yields and production, so the largest operations receive the largest transfers without any finding that their need for stabilization is proportionately larger. Premium subsidies for crop insurance carry no payment limit and no income means test, an asymmetry with the commodity title, where the Agricultural Act of 2014 set a combined Agriculture Risk Coverage and Price Loss Coverage limit of 125,000 dollars per person and an adjusted gross income eligibility limit of 900,000 dollars. Benefits capitalize into land values and rents, transferring a measured share of the appropriation to landowners who may not farm at all and raising entry costs for beginning farmers. Each of those counts accepts, for the sake of the argument, that correlated risk and income volatility justify a public framework. The dispute is over whether the framework as designed sends the money where its stated aim points, and over whether the instruments chosen carry side effects, in land markets, in planting decisions on marginal acres and in trade relations, that a differently designed framework would avoid. On the myths and contested claims that farm subsidy numbers attract, the distinction between attacking a design and attacking risk support itself is the distinction that keeps the argument honest, and it is recorded in the myths and contested claims that farm subsidy numbers attract in the form the debate usually takes.

A final discipline governs the use of every finding in this article, and the counter-reading states it plainly so that the case for the programs is answered on the evidence and not on an extrapolation. No finding reported here is extended to any proposal for future legislation, and no measured effect from a payment structure that operated in a stated period is treated as a prediction about a structure that has not operated. The concentration figures describe payments as the statutes of their period designed them. The capitalization estimates describe rental and land markets as they responded to the payment streams of their period. The cotton findings describe instruments that were repealed, replaced or reformed in the settlement and in the Agricultural Act of 2014. A reader who accepts the strongest case for risk support and accepts every finding in this article holds two compatible positions, because the findings measure what the programs as built actually did, and the case for the programs explains why a nation might build support again with different limits, different targeting and different instruments. The evidence has to answer the case for the programs. It does not have to pretend the case was never made.

The case also answers, in advance, the simplest version of the opposing argument, the claim that transfers of this size must be waste in their entirety because private markets exist for insurance, credit and land. Private markets for each of those functions do exist, and each one prices the correlated character of agricultural risk into its terms, through premiums that load for systemic loss, credit that contracts when collateral values fall with commodity prices, and land markets that transmit every expectation about future support into the price of entry. The existence of a market is not evidence that the market clears at terms a sector can survive through a multi-year trough, and the historical record of disaster legislation is evidence that elected governments, confronting the consequences of letting it fail, intervened afterward on worse terms than a standing framework offered in advance. A serious defense of the programs can concede the concentration figures, the capitalization estimates and the insurance effects on marginal land, and maintain that a redesigned framework, with limits extended across titles and payments aimed at income risk rather than at production scale, would answer the findings without surrendering the stabilization the statutes set out to buy.

What the Record Supports

Measured against the aims the statutes state for themselves, the record divides cleanly between what farm support plainly did and what it plainly did not do. The support plainly stabilized incomes in bad years at scale, delivering payments and insurance indemnities through standing rules to a production base that faced correlated weather and price losses no private market had shown it could carry alone, and the framework did so predictably enough that lenders priced expected support into the credit on which the sector runs. It plainly produced documented conservation gains on retired land, where the Conservation Reserve Program record of prevented erosion, nutrients kept out of waterways, restored wetlands and established habitat stands as the most concrete purchased outcome in the evidence reviewed. And it plainly produced a priced international consequence in cotton, where a formal dispute converted domestic program design into findings of prohibited and actionable subsidies and a settlement path that cost the United States 805.5 million dollars in payments to Brazil, a figure that gives the abstraction of trade distortion a precise and unusual dollar value.

The other half of the assessment is equally plain. Support did not distribute itself in proportion to need, because payments keyed to acres, yields and production delivered more than three quarters of recorded subsidies to the top tenth of recipients across two decades of payment records, a pattern that mirrors production and therefore hardens with every increment of consolidation in the sector. It did not keep the benefit with the operating farmer in full, because rental and land markets transferred a measured share to landowners, capitalized payment value into purchase prices, and converted part of the intended income support into a higher cost of entry for the farmers who followed. And it did not change planting decisions very much through its decoupled instruments, which the production research ranks as the least distorting support design available, while the larger planting effects that the record does contain arrived through subsidized insurance on marginal land, the instrument with the fewest limits attached to it.

That distance also explains why debate about the programs so often talks past itself. A defense built on stabilization points to real income rescued in real trough years, and a critique built on distribution points to real dollars concentrated at the top of the payment record and capitalized into the ground under the farm, and both descriptions fit the same programs in the same years because the design produced both results at once. The statutes promised support against volatility few sectors match. The record shows a system that delivered the stabilization, charged part of its price to landowners, tenants and trading partners in ways the promise never mentioned, and bought its clearest environmental returns where the payment purchased a defined physical outcome, in tons of soil held in place and acres of wetland restored under contract.

The pattern across both halves is the organizing claim with which the evidence began, now stated as a conclusion rather than a preview. A support system built on land, paid by the acre and by the historical record of production, behaves as support for land. Its benefits concentrate where the land and the production concentrate, settle into the value of the land itself, and pass in part to whoever owns that land when the payment stream is capitalized, while its effects on what is planted remain small exactly where the design worked hardest to make them small. None of that record answers, by itself, whether a nation exposed to the risks described in the statutes should support farm income. It answers the narrower question the article set, which is whether the money landed on the aim the law announced. In part it did, in the stabilization of a volatile sector and in conservation outcomes that can be counted in tons and acres. In part it landed beside the aim, on land values, on rents and on the largest operations in the sector, and the distance between the announced aim and the measured landing place is the finding that any honest account of the programs has to carry. For readers assembling the primary sources, studies and settlement documents behind that account, a single organized record makes the next step easier, and researchers can keep your statute notes, citations, and case chronologies together free on VaultBook.

Frequently Asked Questions

Q: Who actually receives farm subsidy payments?

Payments are made to farm operators and other eligible recipients recorded by the United States Department of Agriculture, but the distribution is heavily concentrated. The Environmental Working Group Farm Subsidy Database, built from USDA payment records obtained under the Freedom of Information Act, found that between 1995 and 2016 the top 10 percent of recipients received 77 percent of the subsidies recorded in the database. An earlier release covering 1995 to 2010 found the top 10 percent of subsidized farms collected 76 percent of commodity payments. Payments are keyed to base acres, payment yields and production, so they scale mechanically with operation size, and concentration of payments largely mirrors concentration of production. Roughly 98 percent of farms are family farms, according to the Economic Research Service, so concentration by size is not the same fact as corporate ownership.

Q: Do farm subsidy payments raise land values?

Research answers yes, with the size of the effect contested. Goodwin, Mishra and Ortalo-Magné in 2003 and Roberts, Kirwan and Hopkins in 2003 found that decoupled payments are capitalized into farmland values, meaning that an expected payment stream raises the price a buyer will pay for the land that carries it. The 2003 Report of the Commission on the Application of Payment Limitations for Agriculture estimated that total government payments in the years it examined had increased United States farmland values by 15 to 25 percent. Capitalization carries a consequence for the stated aim of the payments: a buyer who pays a capitalized price has already paid for the expected subsidy stream in the purchase, and so gains little net support from payments that appear, in program records, to flow to that buyer in full.

Q: Do landlords capture farm subsidy money?

Landlords capture a measured share of it through rent. Barrett E. Kirwan, writing in the Journal of Political Economy in 2009, used policy variation to estimate that tenant farmers who rent the land they cultivate capture about 75 percent of a subsidy, while landowners capture about 25 percent through higher rental rates, with rental rates rising by roughly 25 cents for each dollar of subsidy. Theory had predicted that landowners would capture nearly all of it, so the finding left most of the payment with the tenant, but a quarter leaked to owners, and the landlord share rose where rental markets were more competitive. The size of the channel reflects the tenure base: about 39 percent of farmland in the contiguous states was rented, and 80 percent of rented acres belonged to non-operator landlords.

Q: Do farm subsidy programs change what gets planted?

The answer depends on the instrument. Decoupled payments, which are not tied to current prices or plantings, have small measured effects: Goodwin and Mishra in 2006 found only modest effects on acreage and resource allocation, Weber and Key in 2012 found small positive effects on cropland acreage through wealth and credit channels, and Serra, Goodwin and Featherstone in 2011 found a very small elasticity of output with respect to decoupled payments. Comparative analyses by the Organisation for Economic Co-operation and Development ranked decoupled payments as the least production distorting of the major support instruments. Crop insurance premium subsidies show larger effects. Claassen, Carriazo and Cooper at the Economic Research Service estimated in 2011 that premium subsidies shifted land into crop production, with effects concentrated where insurance made cropping marginal land viable.

Q: Did a WTO case find US farm subsidy rules illegal?

The formal record is more precise than the question’s shorthand. In the cotton dispute brought by Brazil, designated DS267, a World Trade Organization panel in 2004 and the Appellate Body in 2005 found that the Step 2 payments and the GSM-102 export credit guarantees were prohibited subsidies. The same proceedings found that the marketing loan and counter-cyclical payments were actionable subsidies that caused serious prejudice to Brazil through significant suppression of world cotton prices. Direct payments also failed the green box decoupling criteria in part, because producers who took them were barred from planting fruits and vegetables on base acres. Prohibited and actionable are distinct legal categories with distinct remedies, and the findings fell into both, rather than into a single ruling that the programs were illegal in general.

Q: Do farm subsidy programs lower grocery prices?

The arithmetic of the food dollar says no in any large measure. The United States Department of Agriculture Economic Research Service Food Dollar Series put the farm share of the consumer food dollar at 14.6 cents in 2017, the lowest figure since the series began in 1993, with roughly 85 cents of each food dollar going to processing, transport, wholesaling, retailing and food service. Because the farm gate receives so small a share, commodity support cannot move grocery bills very far in either direction, and claims that ending subsidies would make groceries expensive fail on the same arithmetic as claims that subsidies make groceries cheap. The standing exception is a small class of programs designed to raise domestic prices, of which the sugar program, built on import restrictions and price supports, is the described example.

Q: Does conservation spending under farm subsidy law work?

On retired land, the documented record is strong, and on working land the answer is genuinely open. The Farm Service Agency reports that the Conservation Reserve Program, created by the Food Security Act of 1985, has prevented more than 9 billion tons of soil erosion since its creation, reduces erosion by roughly 325 million tons per year relative to pre-program levels, keeps more than 600 million pounds of nitrogen and more than 100 million pounds of phosphorus out of waterways each year, and has restored more than 2 million acres of wetlands. For working lands programs such as the Environmental Quality Incentives Program and the Conservation Stewardship Program, Economic Research Service research by Claassen and coauthors found evidence that a share of cost shared practices would likely have been adopted without the payment, so additionality remains an open research question.

Q: What is the strongest case for farm subsidy programs?

The strongest case rests on market failure and irreversibility. Agricultural yield and price risks are correlated across producers in the same year, which is the condition under which private insurance markets struggle with systemic risk and adverse selection. Net farm income swings by tens of billions of dollars between adjacent years, lenders lend against expected support, and the ad hoc disaster bills that preceded standing insurance programs were slower, less predictable and no cheaper. Support also preserves productive capacity and a producer base through price troughs that would otherwise force exits that cannot be reversed when prices recover, which frames food security as a standing national interest. The strongest critique in the literature accepts much of that case and aims instead at design and distribution, at payments scaled to size, premium subsidies without limits or means tests, and benefits capitalizing to landowners.

Q: Do farm subsidies keep small farms in business?

The payment structure works against that outcome more than toward it. Because commodity payments are keyed to base acres, payment yields and production, the amount a farm receives rises with its size, and the concentration record shows the top tenth of recipients taking more than three quarters of recorded subsidies between 1995 and 2016. A small operation receives a small payment in absolute terms, even where the payment matters greatly relative to its income. Economic Research Service consolidation research by James MacDonald and coauthors found that farms with at least 1 million dollars in sales rose from 31 percent of the value of production in 1991 to about 51 percent in 2015, a shift the payment design mirrors rather than offsets. Stabilization may help a small farm survive a bad year, but the distribution of support does not target survival at the small end of the sector.

Q: Why do farm subsidy totals jump in some years?

Totals jump because several of the largest instruments are designed to pay against market conditions rather than at a fixed rate. Agriculture Risk Coverage and Price Loss Coverage, created by the Agricultural Act of 2014, pay when revenue or prices fall below reference levels set in the statute, so a year of low prices triggers payments across enrolled base acres at once, while a year of strong prices triggers few or none. Crop insurance indemnities follow the same logic from the loss side, rising in years of widespread drought, flood or price decline, when correlated losses produce claims across whole regions in a single season. Marketing assistance loans also expand in low price years. The volatility in the totals therefore reflects the volatility the programs were built to answer, since payments that stabilize income must be largest in the years when market income is lowest.

Q: Do farm subsidies push farming onto land that should stay in grass?

Premium subsidies for crop insurance are the instrument the research identifies. Claassen, Carriazo and Cooper at the Economic Research Service estimated in 2011 that premium subsidies shifted land into crop production, with effects modest at national scale but concentrated where insurance made cropping marginal land viable. Related Northern Plains simulation work by Claassen and coauthors found that insurance subsidies materially raised grassland conversion under low price scenarios and mattered little under high prices. The mechanism runs through expected return: an insurance guarantee, supported by a premium subsidy averaging about 62 percent of the premium according to the Congressional Research Service, raises the expected return of cropping risky, drought prone or wet land relative to leaving that land in grass, and prevented planting provisions can add to the same incentive.

Q: Why did the United States pay Brazil rather than change the cotton program at once?

The payments bought time and avoided retaliation while program change waited on legislation. After arbitration in 2009 authorized Brazilian countermeasures with a fixed component of 147.3 million dollars per year plus a variable amount, and Brazil prepared retaliation covering roughly 830 million dollars of trade in goods and services, the Framework Agreement of June 2010 substituted annual payments of 147.3 million dollars to the Brazilian Cotton Institute for technical assistance and capacity building in Brazil’s cotton sector. Cotton support was embedded in farm bill statutes that only Congress could rewrite, and the governing farm bill ran on a multi-year cycle. The United States paid 505.5 million dollars under the framework before payments stopped in October 2013, and the Memorandum of Understanding of October 1, 2014 added a one-time final payment of 300 million dollars, bringing the settlement path total to 805.5 million dollars.

Q: Are crop insurance premium subsidies a form of farm subsidy?

Yes, by any accounting that follows federal cost and producer benefit. The federal government pays a share of the premium that averages about 62 percent of the total premium, according to the Congressional Research Service, and that share is a transfer that lowers the producer’s cost of coverage below the actuarial price the risk would otherwise command. After the Agricultural Risk Protection Act of 2000 raised subsidy rates, participation rose to roughly 90 percent of eligible acres, which indicates how strongly the subsidy drives the market it supports. Premium subsidies also differ in design from commodity payments in a way that matters for distribution: the Agricultural Act of 2014 set a per person limit of 125,000 dollars for Agriculture Risk Coverage and Price Loss Coverage combined, while federal crop insurance premium subsidies carry no payment limit and no income means test.

Q: How do researchers tell whether a payment is really decoupled?

Researchers test the payment against the decisions it is supposed to leave untouched. A genuinely decoupled payment does not vary with current prices, current plantings or current production, so econometric work examines whether acreage, crop choice and output respond to the payment after those formal conditions are met. David A. Hennessy explained in 1998 why a response can occur even under formal decoupling: payments raise wealth, and under decreasing absolute risk aversion a wealthier producer accepts more risk, while stabilized income operates like insurance, so planting can expand through wealth and insurance effects. Goodwin and Mishra in 2006 and Weber and Key in 2012 measured those channels and found small effects. Expectations provide a further test, since the base updating permitted by the 2002 act reconnected planting decisions to expected future payments, as Daniel A. Sumner argued in 2003.

Q: Do conservation payments go to the same large operations as commodity payments?

The programs reach land through different doors, so the overlap is partial rather than complete. The Conservation Reserve Program pays rent on environmentally sensitive cropland retired under contracts of 10 to 15 years, and eligibility follows the land and its erodibility rather than the production scale of the operator, with rental payments running about 1.8 billion dollars annually according to the Farm Service Agency and enrollment at roughly 22 million acres in 2020. Working lands programs such as the Environmental Quality Incentives Program and the Conservation Stewardship Program pay operations that keep land in production, so their recipients are active producers and can include large operations. Commodity payments, by contrast, are keyed to base acres and production history. A large operation can therefore appear in both systems, but a retired acre enrolled by a small owner or a non-operator landlord can receive conservation rent without any corresponding commodity payment.

Q: Would ending farm subsidies reduce food production by very much?

The production research points to a modest aggregate effect, unevenly distributed. Decoupled payments were ranked by the Organisation for Economic Co-operation and Development as the least production distorting of the major support instruments, with an output effect per dollar far below that of market price support or output linked payments, and measured acreage effects for direct payments were small in the work of Goodwin and Mishra and of Weber and Key. The larger effects sit in crop insurance, where Claassen and coauthors found premium subsidies shifting marginal land into production, so the acres most likely to leave cropping without support are the risky, drought prone or wet acres where insurance had made cropping viable. Land values would also adjust, since capitalized support is priced into farmland, and a production base carrying debt against supported income would face a difficult transition even where physical output changed little.

Q: Did the cotton case change how other countries design farm support?

The case supplied a working demonstration of how support design would be judged, and its lessons are visible in the categories it applied. The panel and Appellate Body reports in dispute DS267 distinguished prohibited subsidies, which had to be withdrawn, from actionable subsidies, which had to answer for serious prejudice shown through price suppression in world markets, and that distinction tells any government designing support which instruments carry litigation exposure. The finding that direct payments failed the green box decoupling criteria in part, because planting fruits and vegetables on base acres was barred, showed that a payment labeled decoupled can lose its shelter over a single condition attached to the acres. The case also demonstrated that a mid-sized exporter could sustain a claim against the largest agricultural subsidizer through a decade of proceedings, compliance review and authorized countermeasures, and collect a settlement of 805.5 million dollars.

Q: Why are there no payment limits on crop insurance premium subsidies?

The asymmetry is a matter of statutory design rather than oversight. The Agricultural Act of 2014 set a per person payment limit of 125,000 dollars for Agriculture Risk Coverage and Price Loss Coverage combined and an adjusted gross income eligibility limit of 900,000 dollars, while federal crop insurance premium subsidies carry no payment limit and no income means test. Insurance has traditionally been framed in legislation as a risk management partnership in which the producer pays a share of an actuarial premium, and capping the subsidy would cap the coverage a large operation could buy through the federal system, pushing the largest risks toward private products or self-insurance. Defenders of the design argue that broad participation, which reached roughly 90 percent of eligible acres after subsidy rates rose under the Agricultural Risk Protection Act of 2000, keeps the risk pool sound. The literature treats the missing limits as one of the central design choices under criticism.

Q: How much farmland do tenants farm in the United States?

Tenants farm a large share of the national land base. The Tenure, Ownership, and Transition of Agricultural Land survey of 2014, conducted by the Economic Research Service and the National Agricultural Statistics Service, found that approximately 39 percent of the 911 million acres of farmland in the contiguous 48 states was rented, a total of about 354 million acres, and that more than half of cropland was rented. Of the rented acres, 80 percent, or 283 million acres, was owned by non-operator landlords who do not farm the land themselves, and retired farmers made up 38 percent of those non-operator landlords. That tenure structure is the reason the capitalization findings matter beyond the individual lease: a payment system that operates across a land base farmed substantially by tenants will divide its benefit between operators and owners in proportions set by the rental market.

Q: What would happen to land values if farm subsidies ended?

The capitalization research implies that values would fall by roughly the capitalized value of the expected payment stream, though the estimates carry wide bands. Goodwin, Mishra and Ortalo-Magné in 2003 and Roberts, Kirwan and Hopkins in 2003 established that decoupled payments are capitalized into farmland values, and the 2003 Report of the Commission on the Application of Payment Limitations for Agriculture estimated that total government payments in the years it examined had raised United States farmland values by 15 to 25 percent, which provides an upper guide to the adjustment rather than a forecast. Rental rates would move first and by a measurable share, since Kirwan estimated in 2009 that rents rise by roughly 25 cents for each dollar of subsidy. Owners who bought land at capitalized prices would absorb the loss, which is one reason transitions in support design carry consequences well beyond the farms that receive the payments.