The Appeal — and the Risk — of Data
Data is the most common way to settle disputes in valuation work. It validates claims and reduces pure-opinion noise. Damodaran, however, is careful to flag two persistent limitations practitioners should internalize before leaning too hard on any dataset.
First, data can be biased. Analysts who approach a dataset with prior conclusions can find ways to make the numbers say what they already believed. Transparency about sample construction and statistical methodology is the only real safeguard.
Second, historical data is not a promise about the future. Patterns that held for decades can break — the small-cap and value premiums being two recent examples — and mechanical extrapolation from history is one of the more common errors in practitioner work.
The Sample
For 2025, Damodaran's dataset pulls from S&P Capital IQ, Bloomberg, FRED, NAIC, and Political Risk Services. It covers 47,810 publicly traded firms with available market pricing at year-start.
The geographic and sectoral distribution tells a familiar story: US companies dominate market capitalization globally, and technology remains the most highly valued sector worldwide, with regional variation in sector composition.
The Variables
Industry-level statistics span four familiar categories:
- Risk (betas, volatility, default probabilities)
- Profitability (margins, returns on capital)
- Leverage (debt ratios, coverage)
- Dividends and capital return
These are grouped by their primary use case — corporate finance, valuation, or pricing. Some metrics come from financial statements; others reflect market prices or macro inputs.
The Timing
The dataset is compiled at the start of 2025. Accounting figures come from the most recent quarterly filings, while market-based metrics capture early-year valuations. Accounting ratios tend to be stable across the year; pricing multiples age fast. A user who pulls Damodaran's dataset in Q3 should weight that difference accordingly.
The Estimation Process
Data licensing constraints prevent sharing company-level data, so reporting is aggregated at the industry level. The methodology deals with:
- Sampling bias by including multiple aggregation variants
- Outliers by trimming or using medians where appropriate
- Missing data by reporting statistics only for firms with complete information, which introduces its own selection bias
Every one of those decisions involves trade-offs. No single aggregation method is universally correct — practitioners should understand which variant fits their use case.
Accessing the Data
Current and archived versions of the dataset are published on Damodaran's website and are free to use. He does not accommodate custom requests, but anyone with a data vendor subscription and enough patience can reconstruct the dataset from scratch.
Practitioner Takeaways
- When using industry averages as valuation inputs, read the methodology notes — they affect the numbers more than users typically realize
- Be cautious about extrapolating historical averages into forward assumptions, especially for risk premiums and growth rates
- Cross-check industry averages against a peer set you've hand-selected for comparability
Bottom Line
Damodaran's dataset is one of the most valuable free resources in valuation practice, but it is a tool, not a verdict. Understanding what's in the sample, how variables are estimated, and when the data was snapped are all preconditions for using it well.