A discount for lack of marketability exists because a private-company interest cannot be converted to cash on demand: there is no ready buyer, the eventual holding period until a liquidity event is uncertain, real transaction costs will be incurred when a sale finally happens, and the holder faces genuine information asymmetry relative to a public-market buyer. Three families of evidence anchor a defensible figure. Restricted-stock studies compare a public company's freely-tradable share price against the price of an SEC Rule-144 restricted ("letter") stock issued by the same company at the same time, a clean natural experiment in pure illiquidity, holding the underlying business constant. Modern studies, run against the shorter six-month minimum holding period that has applied since 2008, tend to show meaningfully smaller average discounts than the older studies built around the longer one- to two-year holding periods that applied before 1997, a live reminder that "the" restricted-stock discount is a moving, holding-period-dependent number, not a constant (check against the current study database, the Stout Restricted Stock Study is the modern successor to the earlier FMV Opinions/Management Planning studies). Pre-IPO studies (the Willamette Management Associates/Emory series) instead compare a company's own private placement price shortly before its IPO against the eventual IPO price; these typically show materially wider average discounts, often cited north of 40–60%, but are methodologically contested because some of that gap reflects genuine value created between the placement and the IPO rather than pure illiquidity. Option-pricing proxies, this module works the Chaffe (1993) model, treat the inability to sell for a horizon T as forfeiting a put option: the right to sell at today's price at any moment, which a marketable holder effectively owns for free and an illiquid holder does not. Worked Example 5 computes this precisely, by hand, from Black-Scholes.
Advertisement
What are the Mandelbaum factors for DLOM?
None of the three families hands you a single "correct" number; they hand you a range, and the Mandelbaum factors, drawn from the US Tax Court's decision in Mandelbaum v. Commissioner (T.C. Memo 1995-255, Judge Laro), are the reasoned basis for choosing where inside that range a specific interest belongs: (1) the company's financial-statement strength, (2) its dividend or distribution policy, (3) the nature, history, and industry position of the company, (4) the depth of its management, (5) the amount of control carried by the transferred block, (6) any restrictions on transferability in the governing documents, (7) the expected holding period until a liquidity event, (8) the entity's redemption policies, and (9) the costs of a hypothetical public offering. These factors do not compute a percentage; they are the actual written justification an appraisal must carry to survive scrutiny, a discipline the IRS itself required in principle as far back as Revenue Ruling 59-60 (1959), which first demanded a reasoned, factor-based approach to valuing closely-held stock for tax purposes rather than an asserted number.
{
"skill": "Drafting a defensible DLOM/control-premium memo without letting AI originate the number",
"use": "Have an AI assistant enumerate the Mandelbaum factors against your specific fact pattern and draft the memo's factor-by-factor narrative, while you choose and defend the actual percentage from a cited study or a Chaffe computation you ran yourself.",
"tools": ["Chat assistants — Claude, ChatGPT", "Research/agentic tools for locating current study data (still verify the primary study directly)"],
"prompt": "Here is the fact pattern for a private-company interest I am valuing: <describe the entity, the % stake, control or minority, dividend history, expected holding period, and any transferability restrictions>. Walk through each of the nine Mandelbaum factors for this fact pattern and tell me, for each, whether it points toward a LARGER or SMALLER discount than a typical restricted-stock-study average. Do not give me a final percentage.",
"verify": "This is the corpus's highest AI-over-trust zone: a DLOM or control-premium figure is a judgment call, not a lookup, and an AI will produce a specific, confident-sounding percentage ('30% is standard') with no defensible sourcing whatsoever — a textbook case of AI0.01's 'plausible, not true' applied to a number that decides what someone actually inherits. Trace every cited study to its actual publication, recompute the Chaffe figure yourself in code you can inspect line by line, and never let a chat assistant's stated percentage become the number that appears in a memo you sign (Guardrail 1-5, especially Trace and Recompute).",
"diy": "You must already be able to state the levels-of-value hierarchy, compute a Chaffe DLOM by hand, and name all nine Mandelbaum factors before an AI's draft is a useful starting point rather than a trap.",
"market": "US"
}