Outgrace, home

Reading published cost-of-capital data before you use it

Published datasets differ in return method, universe and coverage. Check those before a published rate goes into a capital budget or a board paper.

A discount rate in a board paper usually arrives with a source line that names a dataset and a year. Nobody asks what the dataset counted. The number then moves into capital budgets, impairment tests and acquisition models, where it does its work quietly for years. Before any published cost-of-capital figure goes into a paper, the reader should know three things: whose returns it measures, how those returns were calculated, and how far the series has changed since it was first published. The public documentation answers all three, if you read it.

Portfolios in the 2x3 Size and book-to-market grid, Kenneth R. French Data Library
6
Portfolios in the 5x5 Size and book-to-market grid, Kenneth R. French Data Library
25
Portfolios in the 10x10 Size and book-to-market grid, Kenneth R. French Data Library
100

What the data shows

The Kenneth R. French Data Library, hosted on the Tuck School of Business site at Dartmouth, publishes US research returns. According to the library, its pages show no industry cost-of-capital or discount-rate estimates. That one statement rules out a common shortcut: taking a sector discount rate from a factor library and assuming it describes your industry. The library is a source of return history and factor series. It is not a source of a finished rate for a business.

Start with the documentation page, not the file. The page tells you the universe, the return basis and the change history in a few lines, and the file tells you none of it. In our experience, many reconciliation problems start with a user who downloaded the numbers and skipped the notes. A simple rule helps: if you cannot say which return method sits behind a number, that number should not go into a board paper.

The method changed at the turn of 2024 and 2025, and the change matters more than most users assume. According to the library, the legacy FIZ files were discontinued after the December 2024 data release. From January 2025, US research returns are built on CRSP Flat File Format 2.0, under the CIZ label. The two methods calculate monthly returns differently. Legacy FIZ monthly returns are month-to-month holding-period returns, with dividends reinvested at month-end. CIZ monthly returns are compounded daily returns, with dividends reinvested on ex-dates. A series that runs across that boundary is not one series. The library links a cross-reference guide to the FIZ and CIZ histories, and it should be read before any two periods are joined.

The factor returns carry a second source of movement. In a December 2023 paper, Fama and French explain how US factor returns are produced. According to the library, the same paper estimates the effect of two process changes, and of five CRSP data-improvement projects, on the average values of the SMB and HML factors. The point for a leader is plain. A size or value premium can shift because the underlying data were cleaned or reprocessed, not because the market repriced small or cheap companies. The library also keeps separate pages on changes in CRSP data and on changes in industry specifications. Check both before you compare one period's factor level with another.

Formats matter more than they look. The three-factor series is offered in monthly, weekly and daily versions, as text and as CSV files. The five-factor model appears in its 2x3 form, listed as new, with a daily version as well. According to the library, the five-factor family also appears in 2x2 and 2x2x2x2 forms, both listed as new. Many portfolio sets are offered ex-dividends and as daily files, and some carry Historical Archives links. Choose the frequency your model needs before you download, and record the variant and return basis you used. A model that mixes daily factors with monthly company returns will produce figures that look sound and are not comparable.

The library sorts stocks into portfolios on size, book-to-market, operating profitability and investment. Size and book-to-market grids come in 6, 25 and 100 portfolios, in 2x3, 5x5 and 10x10 layouts. Size and operating profitability, and size and investment, are offered in grids of 6, 25 and 100 portfolios. The book-to-market and operating profitability grid is a 5x5 layout of 25 portfolios. Grid choice is a judgement about how finely you want to separate companies, and a finer grid gives more groups with fewer companies in each. Pick the grid for a reason you can state, and do not pick it for the output that looks most favourable.

Figures in the sources

Each figure as stated on the publisher's page named beside it. The portfolio counts describe grid sizes, not estimates of return or risk.

Developing countries' external debt at a record, International Debt Report 2026, World Ban
$9.4 trillion
Portfolios in the 2x3 Size and book-to-market grid, Kenneth R. French Data Library
6
Portfolios in the 5x5 Size and book-to-market grid, Kenneth R. French Data Library
25
Portfolios in the 10x10 Size and book-to-market grid, Kenneth R. French Data Library
100

Source: World Bank; mba.tuck.dartmouth.edu

What it means for a leader

Most leaders will not need a factor library at all. They need a rate that matches a specific business, in a specific currency, at a specific date. The French data cannot supply that. It describes the premiums that US listed shares have produced, as the library measures them, over the periods it covers. Use it, if you use it, as one input to a model you can explain, and say so in the paper. A director who asks where the premium came from should get an answer that points to a document, not to a habit.

Regional differences cannot be read off the sources in this article. The fact sheet holds no region-level cost-of-capital figure for the euro area, for Europe or for any developing market. What it does offer is context. According to the World Bank's International Debt Report 2026, developing countries' external debt reached a record $9.4 trillion. The same report says private creditors are back, bond markets have reopened, and borrowing costs sit at decades-high levels. Those are sovereign and external borrowing conditions. They tell a director something about the cost of debt for some borrowers in some markets. They do not give a firm's discount rate, and they should not be used as one.

The European Central Bank points in a related direction. Its Vice-President, Boris Vujčić, has said that Europe needs more market-based finance to support investment and economic growth. The ECB notes that non-banks play an increasingly important role in the financial system. Its risk analysis covers AI valuations, fiscal imbalances, non-bank vulnerabilities and cybersecurity threats, and it describes the geoeconomic environment as highly uncertain. A European firm reading that list should ask a plain question: does the rate in the paper reflect the market the company actually raises money in, on the date it was set?

Ownership matters as much as arithmetic. Decide who in the business signs off the rate, who keeps the reconciliation, and when the rate is refreshed. A rate set once in a strategy paper and reused for several budget cycles will drift from the market it claims to describe. The fix is a dated note in the paper stating the source, the version and the review date, with a named person who can answer questions about it at the board.

Adjusting a published rate for your own risk starts with a short reconciliation. Write down five things: the universe the series measures, the return basis, the data frequency, the date of the data and the year of the latest methodology change. Then compare each one with your project or company. If your business is smaller, more concentrated or more heavily borrowed than the portfolios the series describes, the gap is a risk judgement, and the paper should present it as one. The adjustment that holds up is a visible premium with a stated reason. An unexplained change to the headline number does not hold up.

Sector is the weakest link in the chain. The French library sorts by characteristics such as size and book-to-market, not by industry, and it shows no industry cost-of-capital estimates. An industry rate therefore has to come from another publisher, and its documentation needs the same care: the industry definitions, the universe of companies, the period covered and any changes to the classification. The library's own separate page on changes in industry specifications is a reminder that classifications move. A rate built on an old set of industry boundaries may describe a grouping that no longer matches your business.

A practical test helps before you adjust anything. Ask whether a reasonable director could reproduce your rate from the documents in the paper alone. If the answer is no, the number is a conclusion without a route to it. If the answer is yes, the adjustment is open to challenge, and that is the point of putting it on paper.

Limits of the data

The sources here are narrow. The French library covers US research returns only. Its factor and portfolio series are research measures, not firm-specific discount rates and not industry cost-of-capital estimates. The estimates in the December 2023 paper, of how process changes affected the SMB and HML values, are the authors' findings about the series. This article has not re-run them. The World Bank and ECB material describes conditions in debt and financial markets. It is not a published cost of capital for any sector or region. We have not verified any industry or regional dataset, and this piece does not name one. Before a rate goes into a board paper, a valuation specialist or the company's auditor should review the method. Where a decision carries tax or legal consequences, the relevant adviser should be consulted first.

Newsletter

The Briefing

One email a week: the leaders we profiled, the ideas worth stealing, and nothing else.

Weekly, free, one click to leave. Read our privacy policy.