Lesson 1 of 8
What the first studies measured
Before arguing about whether the value premium is real, read what was measured: which market, which years, which ratio, and how the money was weighted.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- A value premium is not one number. It is a number produced by a universe, a period, a ratio and a weighting rule, and changing any of the four changes it.
- Most of the founding results are equal weighted, which puts as much money in the smallest company as in the largest.
- The longest public record of the factor, running from July 1926, compounds at about 3.6 percent a year, and almost none of the decades look like that average.
Nobody measured "value". They measured portfolios built by a stated rule over a stated period, and then wrote down the difference. Reading the rule is the difference between knowing a result and repeating one.
The high minus low book-to-market factor, by decade
- 1940s9.55%
- 1970s7.65%
- 2000s7.38%
- 1980s5.59%
- 1960s3.23%
- 1990s−0.35%
- 2010s−2.62%
Four choices that make a number
Read these before reading any premium
-
The universe. NYSE only, or NYSE with AMEX, or all three including NASDAQ. Adding NASDAQ adds thousands of small companies and changes the answer.
-
The period. July 1963 is where the usable accounting data starts in these files, so most of the famous results begin there rather than at the start of the price record.
-
The ratio. Price to book, price to earnings, price to cash flow and dividend yield sort the market differently, and a company cheap on one is often not cheap on another.
-
The weighting. Equal weighted gives every company the same dollars. Value weighted gives each company its share of the market. The two answers differ by points, not decimals.
The weighting choice, made concrete
Tweedy, Browne's collection of studies prints the construction for each one it tabulates. The study covering July 1963 to December 1990 ranked all NYSE, AMEX and NASDAQ companies in the price research file with accounting data available, sorted them into price-to-book deciles, then sorted each decile again by market capitalisation. The returns are average annual and equal weighted.
That last phrase is not a footnote. Equal weighting puts the same money into a company worth 40 million dollars as into one worth 400 billion. It is a defensible way to measure whether a characteristic predicts returns. It is not a description of a portfolio anyone can hold at size, which is the subject of a later lesson in this course.
The relation is not a straight line
What the long record looks like
The factor series behind the figure above starts in July 1926 and runs to July 2026 in the file downloaded for this lesson. Compounded across the whole century, the high minus low book-to-market return is about 3.59 percent a year. From July 1963, where most of the famous studies begin, it is about 3.10 percent.
Neither figure describes any decade in the chart. Two of the ten decades are negative. That is what an average across a century buys you: a summary that almost never happened.
Check yourself
Why does the weighting rule matter to a measured value premium?
- It changes the tax treatment
- Equal weighting puts as much money in the smallest company as in the largest, which a large portfolio cannot do
- It determines the discount rate
Show the answer
Answer: B.
Equal weighting is a legitimate way to test whether a characteristic predicts returns, and it is not a portfolio that can be held at size. The two are different claims.In the table above, which decile produced the highest first-year return?
- The cheapest decile, number 10
- The ninth decile
- The most expensive decile, number 1
Show the answer
Answer: B.
The ninth decile returned 18.3 percent against the cheapest decile's 17.3 percent. The relation runs in one direction without being monotonic at the end.The book-to-market factor compounded at about 3.6 percent a year from July 1926. How many of the decades shown resemble that figure?
- All of them
- Almost none, and two are negative
- Exactly half
Show the answer
Answer: B.
The decade returns run from 9.55 percent to minus 2.62 percent. A century-long average is a summary of periods that mostly look nothing like it.Sources
- Fama and French three-factor monthly series, Kenneth R. French data library, file built from the July 2026 CRSP database, downloaded and compounded for this lesson on 2026-09-13.
- Tweedy, Browne, What Has Worked in Investing, 1992, revised 2009: Table 6 for the July 1963 to December 1990 construction and Table 7 for the April 1968 to April 1990 decile returns.
What this leaves out
- The factor series is a long-short research portfolio, not a fund. It buys high book-to-market companies and sells low ones, with no costs, no taxes and no borrowing limit.
- Decade figures are computed here by compounding monthly returns inside calendar decades, and the 2020s are excluded from the chart because that decade is incomplete.
- The decile table is the booklet's own tabulation, not this site's computation.
- None of it is a statement about any company, and none of it is advice.