Two portfolios can both be honestly called value, be cut from the same universe on the same date, and be built to the same size, and still hold seven names in ten differently. This paper measures that overlap across 347 four week dates and traces it to the thirteen construction decisions that sit between a set of financial statements and a value ranking, almost none of which the profession publishes. What follows is a finding about disclosure rather than about markets, and an argument about how a factor is built, not a view on any individual security.
Abstract
Two portfolios, both honestly called value, cut from the same universe on the same date and to the same size, hold seven names in ten differently. The reference book ranks a composite of four accounting ratios inside each company’s sector. The comparison book ranks book to market pooled. The three in ten they share carry 13.31 percent of the reference book’s market capitalization. Both shares are medians over 347 four week dates, 3 January 2000 to 13 July 2026. That is not a finding about markets. It is a finding about disclosure.
Thirteen decisions stand between a set of financial statements and a value ranking. The profession publishes almost none of them. Each is taken with the argument on both sides, written down before any return was seen, and priced against the alternative afterward. Three of the thirteen measured worse than the alternative. All three are printed at full size.
All results are gross. Two things are conceded before any evidence and neither is qualified later. The accounting component of market to book is unpriced. A value premium is earned where no book value exists. The claim left is narrower: accounting choices nobody publishes decide which companies a value rule holds. This paper publishes its own, including where its construction did not describe itself.