The Dollar-for-Dollar Test for Retained Earnings

Definition

Buffett splits corporate earnings into two categories: "restricted" earnings, which must be reinvested just to hold the business's competitive position, sales volume, and financial strength, and "unrestricted" earnings, which could equally well be paid out or kept. For the unrestricted portion, he proposes a single test to decide: retain the money only when there's a reasonable, evidence-backed prospect that every dollar kept will create at least a dollar of market value for owners — meaning the capital retained must earn a return at or above what shareholders could get investing it themselves.

In the Book

The 1984 letter's "Dividend Policy" essay introduces the concept by pointing out that most companies state a payout policy — "40% to 50% of earnings" — with no analysis of why that ratio serves owners at all. Buffett illustrates restricted earnings with a real, ironic case: Consolidated Edison, whose stock traded at one-fourth of book value under punitive regulation, meaning every retained dollar became only 25 cents of market value — yet the company kept reinvesting anyway while construction signs around the city proclaimed "Dig We Must." Against that, he lays out the dollar-for-dollar standard for unrestricted earnings as the only legitimate reason to retain rather than distribute them, explicitly rejecting the common motives of empire-building or financial comfort that lead managers to hoard cash regardless of what it can earn.

Why It Matters

This turns a governance question that's usually settled by convention or manager preference — "should we retain or distribute" — into an arithmetic test anyone can apply and audit after the fact: did the capital kept actually create value, or did it just accumulate? It exposes a subtle trap where an institution keeps doing something (reinvesting, expanding, staffing up) because that's the default, without ever checking whether the marginal dollar spent that way beats the next-best alternative use. The same test generalizes to any allocator sitting on discretionary resources — time, headcount, budget — where the honest question is never "can we find a use for this" but "does this use beat what the resource could do elsewhere."