The packet is thick, and the page that matters is near the back. In front are the photographs: the founder on a stage, the couple who paid off the house, the car with the company crest on the grille, the flip-chart arithmetic that turns a few hours a week into a second income and then a first. The page near the back is plainer. It reports what the people already in the business earned last year. Most of them earned almost nothing. The two halves of the packet do not contradict each other. The photographs are real; those people exist. The table is real; that distribution is real. What separates them is which one the buyer uses to estimate what they themselves are about to earn.
The same packet, in different clothes, is handed across a desk at a for-profit college, bound into a coding bootcamp’s enrollment agreement, and filed as the disclosure document a franchise gives a prospect. Each sells entry into an uncertain income opportunity, and in each the buyer has to estimate a return from evidence the seller assembled. The legal regimes differ sharply. An MLM’s earnings claims are governed by the law against deceptive advertising; a for-profit college’s are policed through a test that decides whether its programs stay eligible for federal money; a bootcamp’s income-share agreement is governed by the law of lending; a franchise’s earnings claims are governed by a rule that lets the seller decline to make any. What runs underneath the four is a single recurring dispute: who must produce, define, and substantiate the numbers a buyer uses to price a future.
Chapter 2 examined businesses priced by enforcement, where the governing question was the odds of being caught. This chapter examines markets where the governing dispute is over the evidence of likely outcomes: who holds it, who must produce it, and in what form it reaches the buyer before the sale. The distinction marks a different failure mode. A fraud hides a fact. These markets can disclose every fact and still leave the buyer estimating from a number the seller chose, framed, and, in one of the four, is not required to produce at all.
Deception law and the earnings claim
Multi-level marketing’s earnings claims are governed by the ordinary law against deceptive advertising. In 1979 the Federal Trade Commission examined Amway and held that its particular plan, on the record before it, was not an illegal pyramid. The line the Commission drew, following its earlier Koscot decision, ran between a plan that pays chiefly to recruit new participants and one anchored to genuine sales to real customers. Amway’s plan qualified on that record because it charged no substantial fee to join, required a sponsor to buy back a departing recruit’s unsold inventory, required that a set share of product actually be resold, and required each distributor to make retail sales to ten different customers a month. The same order separately restricted the company’s earnings claims: it barred misrepresenting how much distributors were likely to make and required that any above-average earnings claim be accompanied by disclosure of either the recent-year average or the percentage of distributors who achieved the represented figure. The decision resolved one company’s plan on one record. It did not rule multi-level marketing lawful as a category, and it did not require the annual income statements some companies now publish.
Where a company does publish one, the statement shows a distribution with almost everyone near the bottom. Amway reported a mean 2024 income of $723, before expenses, among all its United States distributors at the Founders Platinum level and below. The figure is a mean rather than a median, and because the distribution is right-skewed the mean sits above the typical case; it is measured before the distributor’s own costs and counts every distributor at those levels, including those who reported no product sales. The number is not concealed. It is disclosed, accurate, and framed by the party using it to recruit, which is the recurring feature this chapter follows rather than any single figure.
Deception law sets a boundary to the earnings claim, and Herbalife crossed it. In its 2016 complaint the Federal Trade Commission estimated that, even assuming distributors resold 75 percent of the product they bought at full suggested retail price and counting no expenses other than their payments to Herbalife, half of Herbalife’s “Sales Leaders,” its distributors at the Supervisor level and above, averaged under $5 a month in profit from retail sales. Herbalife settled the charges without admitting or denying them, paying $200 million and restructuring its compensation. The case is not evidence that the model stays lawful. It is an instance in which controlled framing crossed into a misrepresentation the law reached, which is how the boundary becomes visible.
A test of eligibility
For-profit colleges sell a projected return in the form of a credential, and the money that buys it is mostly federal. The law lets a for-profit school draw up to 90 percent of its revenue from federal student aid under Title IV of the Higher Education Act, a cap set at 85 percent in 1992 and loosened to 90 in 1998. Benefits from the GI Bill and the Defense Department’s tuition assistance did not count as Title IV, so each dollar of that money, placed on the non-Title-IV side of the ratio, could support as much as nine dollars of Title IV revenue while a school stayed within the cap. The arithmetic is a revenue ratio, not a head count, and it gave schools a reason to recruit veterans until a 2021 law folded those benefits onto the federal side, effective for the institutional fiscal years that began in 2023.
The evidence dispute in this market has a name, gainful employment, and it is a test of the very number a buyer is trying to estimate. The rule measures a program’s graduates’ debt against their earnings and, in its current form, against what a worker with only a high-school diploma earns, and it withdraws federal aid from programs that fail. The stated concern, programs whose graduates’ outcomes do not justify the federal money, has persisted across administrations, while the machinery answering it has been issued, struck down, rebuilt, rescinded, and revived. That sequence is the second figure’s subject.
The measured return is small. Comparing for-profit certificate students with matched students at public institutions, Cellini and Turner found the for-profit students earned about $2,100 less per year after attendance, combining lower pay and lower employment; against observably similar people who did not enroll at all, the earnings gain could not be distinguished from zero. A back-of-the-envelope calculation by the authors, sensitive to its assumptions about debt and the length of a working life, leaves the average certificate student modestly worse off once interest is counted. Program-level data widen the frame: an analysis of the Department of Education’s College Scorecard, spanning institutions of every sector, found that at about 30 percent of the institutions studied, more than half of former students earned less than a typical high-school graduate ten years after they first enrolled, an outcome the analysis suggests may partly reflect low completion rates.
The law of lending
The income-share agreement is the newest instrument and a direct test of who must produce the evidence. A coding bootcamp trains a student and, in place of tuition, takes a fixed share of the student’s income for a set period once earnings pass a threshold. The design shifts part of the downside toward the school, and the pitch is that the school is therefore paid only when the student is. The number that makes the trade look sound, the placement rate and the salary that follows, is still compiled and defined by the school. BloomTech, the bootcamp once called Lambda School, marks the boundary. In 2024 the Consumer Financial Protection Bureau found that it had claimed job-placement rates as high as 86 percent during 2017 to 2019, advertised at least 71 percent from 2019 onward, reported placement to its own investors consistently nearer 50 percent, and placed its 2020 graduates, measured in the first half of 2021, at approximately 30 percent. The Bureau also found the school had told students its income-share agreement was not a loan and carried no finance charge, when the agreements carried a finance charge averaging near $4,000 among students who completed repayment, and a single missed payment could accelerate the full $30,000 cap. As with Herbalife, the finding was that the framing had crossed into a misrepresentation the law reached.
The Bureau’s underlying determination was that the instrument was not new to the law. An income-share agreement is credit, a loan subject to the disclosures of the Truth in Lending Act, a point the Bureau had already made against another provider, Better Future Forward, in 2021, and that California’s financial regulator made the same year in a licensing consent order with the servicer Meratas. The lending regime governs the earnings estimate in this market by governing the loan that finances it, which is why the dispute here is fought over disclosures a lender must give rather than over an advertising standard.
An adjacent case
Franchising sits next to this group rather than inside it, and the difference is worth stating precisely. A franchise sells an operating asset, a trademark, a system, a territory, and a product customers already want, and not only an estimate of a return. Its earnings evidence is governed by the Federal Trade Commission’s Franchise Rule, which requires a long disclosure document but makes the financial performance representation, Item 19, optional. A franchisor may present earnings figures, with a reasonable basis and written substantiation, or decline and print a statement that it makes no representation about financial performance, and, apart from narrow exceptions the rule itself states, it forbids performance claims made outside that document. The element franchising shares with the rest is narrow and real: the franchisor holds system-wide performance data a prospect cannot independently reproduce. When the Government Accountability Office examined the Small Business Administration loans made to one franchise brand’s franchisees between 2000 and 2011, it found 74 of 170 had defaulted, with four lenders accounting for 55 of the defaults; the review covered one organization selected for scrutiny, not an industry sample, and a loan default is not the same event as a participant’s net loss. The resemblance is in who holds the evidence, not in how often the buyer is left worse off.
Who must produce the evidence
Set the four side by side and the common thread is not a shared revenue model or a shared legality. It is a shared question about evidence. In each market the buyer is estimating a future return, the seller holds the data that would inform the estimate, and the legal fights keep landing on the evidence: what must be shown, by whom, and in what form. Where the law’s instrument is disclosure, it is a weaker one than prohibition, and its weakness is specific rather than moral. Disclosure can leave substantial room for selective framing, especially when the outcome distribution is skewed, so that a mean stands in for a typical case, and when the buyer cannot independently verify the base rates the seller reports. Where the framing crosses into a false statement, as with Herbalife and BloomTech, another body of law is available. Where it does not, the buyer is left with a true number, chosen and arranged by the party selling the future it describes.
Multi-level marketing produces pyramid-shaped earnings distributions.
Confirmed Moderate confidence
The claim is about the shape of the distribution, and the disclosures show that shape directly. Amway reported a mean 2024 income of $723, before expenses, among all its United States distributors at the Founders Platinum level and below; the Federal Trade Commission estimated in 2016 that half of Herbalife’s Sales Leaders averaged under $5 a month in retail profit. The pattern is not confined to those two companies: in an August 2024 staff report the Federal Trade Commission reviewed the income disclosure statements of 70 multi-level marketing companies and found that many participants received no payments and the vast majority received $1,000 or less for the year, with nearly every disclosure showing most participants in the lowest two or three ranks; Amway’s own table puts the median for the top half of its paid distributors at $539, against the $723 mean. A distribution in which most participants cluster near the bottom while a thin top tier pulls the mean well above the median is pyramid-shaped in the descriptive sense the claim intends. What the evidence does not establish is the adjacent legal claim, that multi-level marketing is an illegal pyramid scheme. Legality does not turn on the shape of the earnings distribution. It turns on whether compensation is driven by recruiting participants or by genuine sales to ultimate users, the structural test set out in the Commission’s Koscot and Amway decisions, and a plan can produce a steeply unequal distribution and remain lawful. Participant profitability and legal status are separate questions, and the absence of an enforcement action against a given company is not an adjudication that it is lawful. The disclosures confirm the shape; they neither prove nor disprove the illegal-pyramid claim.
Sources
- The shape, from the sellers’ own disclosures and a government estimate: Amway U.S. Income Disclosure (2024) — mean annual income about $723, before expenses, among all U.S. IBOs at the Founders Platinum level and below, including those who reported no sales (a mean, not a median; the distribution is right-skewed). Federal Trade Commission complaint and $200 million settlement with Herbalife (July 2016) — a complaint estimate, settled without admission, that half of Herbalife’s “Sales Leaders” averaged under $5 a month in profit from retail sales.
- The industry-wide pattern: Federal Trade Commission staff report, “Multi-Level Marketing Income Disclosure Statements” (August 2024) — a review of 70 companies’ income disclosure statements, taken at face value: many participants received no payments, the vast majority received $1,000 or less for the year, and nearly every disclosure shows most participants in the lowest two or three ranks. Amway’s 2024 disclosure separately reports, among distributors who received any payment, a top-10% average of $14,251 and a top-half median of $539.
- The legal test, kept separate from the distribution: In re Amway Corp., 93 F.T.C. 618 (1979), applying the recruitment-versus-retail standard of In re Koscot Interplanetary (1975). Legality turns on compensation structure and genuine ultimate-user demand, not on how unequal the earnings distribution is.
- Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Directness is high for the shape: the figures are companies’ own disclosures, a government estimate, and a government review of 70 such disclosures. The binding, weaker leg is the review’s own caveats: the disclosures were taken at face value, none accounts for all participant expenses, and the terms are defined differently by each company, so a full net-of-cost census of participant outcomes across the industry still does not exist; the descriptive “pyramid-shaped” claim is well supported, but not with the completeness that would license high confidence.
What the ledger records
Return to the page near the back. Nothing on it is hidden: the distribution is disclosed, the mean is accurate, the footnote is true. It is the number the buyer needs, produced by the party selling the future it describes. The dispute this chapter has followed is over how much more than that a buyer is owed before signing, and who owes it: who must turn a true mean into a typical case, who must show the base rate behind a placement rate, who must state a projected return at all. The four markets answer through four different bodies of law, and where the answer is weak the burden of the estimate stays with the person least equipped to check it.