# Why a stock's beta differs from site to site

One stock, several betas. Window, return frequency, benchmark index, currency, Blume adjustment and leverage, and what each source says it uses.

*https://stockmarketstack.com/guides/why-beta-differs · background to Fundamental Data & Stock Research Platforms*

**Answer:** Because beta is the slope of a regression, and every input to that regression is a choice: how many years of returns, daily, weekly or monthly, against which index, in which currency, and whether the raw slope is then pulled toward 1. Yahoo labels its figure five years of monthly returns; Value Line uses five years of weekly returns against the NYSE Composite, adjusted; Interactive Brokers defaults to two years of daily returns against the S&P 500.

## How it works

Beta is the slope of a straight line. Take a stock's returns and a market index's returns over the
same periods, fit a least-squares line through the pairs, and the slope is beta: how far the stock
has tended to move, on average, for each 1% move in the index. A beta of 1.2 means 1.2% per 1% of
index move, historically and on average.

Nothing in that definition fixes the inputs. Every source chooses:

- **The window** — how many years of history go into the regression.
- **The frequency** — daily, weekly or monthly returns.
- **The benchmark** — the S&P 500, the NYSE Composite, a local index, a regional MSCI index.
- **The currency** — in which the stock's and the index's returns are measured.
- **The adjustment** — whether the raw slope is published as is, or pulled toward 1.
- **The leverage** — whether the figure describes the shares as they trade, or the business with
  its debt taken out.

Change any one and the slope changes. Two sites that disagree on a stock's beta have usually made
different choices, not different mistakes.

## What each source says it uses

From each source's own pages, read on 9 October 2026. *Not stated* means the page does not say.

| Source | Window | Frequency | Benchmark | Adjustment |
|---|---|---|---|---|
| Yahoo Finance, "Beta (5Y Monthly)" | 5 years | Monthly | Not stated | Not stated |
| [Stock Analysis](https://stockmarketstack.com/tools/stockanalysis), "Beta (5Y)" | 5 years | Not stated | Not stated | Not stated |
| [Koyfin](https://stockmarketstack.com/tools/koyfin), watchlist and screener | 60 months, or as few as 24 for a short history | Monthly, month-end | S&P 500 for US stocks, S&P/TSX Composite for Canada, MSCI EAFE for developed markets outside North America, MSCI Emerging Markets for the rest | "Unadjusted", and "levered" |
| Koyfin, beta chart | 1 year (252 days) by default, user-set | Not stated | User-set | Not stated |
| Finviz ([Finviz Elite](https://stockmarketstack.com/tools/finviz-elite) is its paid tier), "Beta (5 Years)" | 60 months | Monthly | "The index", not named | Not stated |
| [Financial Modeling Prep](https://stockmarketstack.com/tools/financial-modeling-prep), `beta` in the company profile | Not stated | Not stated | Not stated | Not stated |
| Value Line | 5 years, 2 at minimum | Weekly | NYSE Composite | Adjusted toward 1.00, weights not published |
| Interactive Brokers, Risk Navigator | 2 years by default, user-set | Daily | S&P 500 by default, user-set | User can overwrite the value |
| Bloomberg, per a university library guide | 2 years by default | Weekly | S&P 500 by default, for a US stock | Raw and adjusted both shown |
| Aswath Damodaran's sector datasets | 2 and 5 years, blended | Weekly | NYSE Composite for US firms, the local index elsewhere | Weighted blend, see below |

Fidelity's research glossary describes a stock beta measured against the S&P 500 over "the last 36
or 60 months", without saying which applies where.

Bloomberg is the one row not taken from the publisher: its documentation sits on the terminal, and
the figures here are from Brigham Young University's library guide to it.

## Window and frequency

**The window decides which company you are measuring.** Five years of monthly returns covers a
business as it was five years ago; one year of daily returns covers this year's. A company that
has changed its debt load, its business mix or its size inside the window has a beta that averages
over all of those versions of itself.

**The frequency decides how many points the line is fitted through.** Sixty monthly returns,
about 260 weekly ones, about 500 daily ones over two years. More points make the slope less
sensitive to any one month, but daily returns bring their own problem: a stock that trades in
another time zone, or trades thinly, does not move in the same hours as the index, so its daily
returns line up with the index's poorly and the slope is distorted — the problem Elroy Dimson's
1979 paper in the *Journal of Financial Economics* addressed for infrequently traded shares. Weekly
and monthly returns leave more time for prices to catch up, so the mismatch weighs less.

**Short histories are handled differently too.** Koyfin drops to as few as 24 monthly returns;
Value Line to two years of weekly ones; Damodaran does not estimate a beta at all with less than two
years of data. A recently listed company can have a beta on one site and none on another.

## Benchmark and currency

Beta is always relative to something. Against the S&P 500, the NYSE Composite or a local index,
the same stock produces different slopes, because the indices themselves move differently.

The benchmark question is sharpest outside the US. Koyfin measures a European or Japanese stock
against the MSCI EAFE, a regional index. Damodaran measures each non-US firm against "the local
index (generally the most widely followed index in that market - CAC in France, Sensex in India and
Bovespa in Brazil)". The two describe different relationships: sensitivity to a broad developed
market, or to the home market.

Currency is folded into the same choice. A stock's returns in its trading currency against an index
in another currency include the exchange rate's moves, so a London stock's beta against a dollar
index partly measures sterling against the dollar. Yahoo shows the issue without resolving it. On
9 October 2026 it printed a beta of −0.19 for Shell on both its London line and its New York
listing, the same unrounded −0.192 on both pages — one figure shared across two listings priced in
two currencies, with no statement of which index or currency produced it.

## Raw and adjusted beta

In 1971 Marshall Blume reported in *The Journal of Finance* that betas measured over one period
tended to be closer to 1 in the next: high betas came down and low betas came up. An **adjusted
beta** builds that tendency in by pulling the raw regression slope part of the way toward 1.

Weights differ by source, where they are published at all:

- **Bloomberg**, as the BYU library guide reports it: adjusted beta = 0.67 × raw beta + 0.33 × 1.0,
  shown next to the raw figure.
- **Value Line**: "The Betas are adjusted for their long-term tendency to converge toward 1.00",
  with no weights given.
- **Damodaran**: a different blend — two thirds of a two-year regression beta plus one third of a
  five-year one, a missing five-year beta replaced with 1, and "an aggregate check to ensure that
  the global average across all the companies is close to one".
- **Koyfin**: none; its betas are "unadjusted".

The adjustment matters least near 1 and most far from it. Applying Bloomberg's weights to Yahoo's
raw-looking figures from 9 October 2026: Apple's 1.07 would become 1.05; Coca-Cola's 0.32, 0.54;
Shell's −0.19, 0.20. A screen for "beta below 0.5" includes Coca-Cola on one basis and excludes it
on the other. Yahoo does not say whether its figure is adjusted, so this arithmetic only shows the
size of the effect; it is not a claim about Yahoo's method.

## Levered and unlevered beta

A beta measured from share prices is **levered**: it describes the equity, and equity in a company
with more debt swings harder for the same business. Koyfin says so outright; the BYU guide says
Bloomberg's beta is "leveraged" when the firm has carried long-term debt for the past two fiscal
years.

An **unlevered** beta takes the debt back out. Damodaran's definition: unlevered beta = beta / (1 +
(1 − tax rate) × debt-to-equity ratio). His US sector table, with data as of January 2026, shows
what that does where debt is heavy:

| Sector | Beta | Debt/equity | Unlevered beta |
|---|---|---|---|
| Air transport | 1.19 | 91.17% | 0.70 |
| Bank (money center) | 0.76 | 164.19% | 0.34 |
| Computers/peripherals | 1.35 | 4.62% | 1.31 |

Unlevered betas are usually published for sectors, for use in valuation models, rather than on a
stock's quote page. A quote page that does not say is showing a levered figure; a number from a
valuation model or a sector table may not be.

## When the sites agree

On 9 October 2026 Apple's beta was 1.07 on Yahoo ("Beta (5Y Monthly)") and on Stock Analysis ("Beta
(5Y)"), and 1.08 on Finviz ("Beta" on the quote page, "Beta (5 Years)" in its help). Yahoo and Stock
Analysis carried the same unrounded 1.069 in their page data, which points to one upstream
calculation. Finviz documents 60 monthly returns against an unnamed index, close enough to the
others' labels that a one-hundredth gap is what you would expect.

That agreement is the point. When the window, frequency and benchmark match, a large US stock's beta
is much the same everywhere. The large disagreements come from comparing a five-year monthly figure
with a two-year weekly or daily one, a raw slope with an adjusted one, or a local-index beta with a
regional or US one.

## What you can do about it

**Read the label before the number.** "5Y Monthly", "Beta (5Y)", "60 month regression" — the label is
often the only documentation there is. If there is no label, the window and frequency are unknown,
and the figure should not be set against another source's.

**Compare like with like.** Line up window, frequency, benchmark and adjustment before comparing two
betas. Where one source is adjusted and the other raw, apply the published weights to the raw
figure, or undo them on the adjusted one, and say so.

**For a non-US stock, find the index and the currency.** A beta against a regional or US index and a
beta against the home index answer different questions; neither is wrong. If the source does not
name either, treat the figure as unexplained.

**Check whether a beta is the same across listings.** If a London line and a New York listing show
identical betas, one figure is being shared, and at most one of them reflects that listing's own
prices.

**Compute it yourself when the inputs matter.** Beta is a regression you can run: pick the window,
the frequency and the index, align the dates, and record each choice with the result. Interactive
Brokers lets you set the window and the reference index in Risk Navigator; Koyfin lets you set the
window on its chart. Prices for your own regression come from any source in
[market data APIs](https://stockmarketstack.com/categories/market-data-apis), adjusted for splits and dividends as
[why adjusted close differs](https://stockmarketstack.com/guides/why-adjusted-close-differs) explains.

**Record which beta you used.** Window, frequency, benchmark, adjustment, leverage and the date you
read it, kept with the number.

## Tools this bears on

- [Koyfin](https://stockmarketstack.com/tools/koyfin.md) — A browser research terminal for global equities, funds, macro and estimates.
- [Stock Analysis](https://stockmarketstack.com/tools/stockanalysis.md) — Five years of financials and a working screener, free and without an account.
- [Finviz Elite](https://stockmarketstack.com/tools/finviz-elite.md) — Real-time quotes, advanced filters and CSV export on top of the free Finviz screener.
- [Financial Modeling Prep](https://stockmarketstack.com/tools/financial-modeling-prep.md) — Financial statements, ratios and filings for 70,000+ securities across 60+ exchanges.

## FAQ

### What does Beta (5Y Monthly) on Yahoo Finance mean?

It is the label Yahoo prints next to the figure, in the Stock Price History block of its statistics page, and it says the regression used five years of monthly returns. Yahoo does not say, on that page or in a help article we could find on 9 October 2026, which index the returns were measured against, whether the figure is adjusted toward 1, or who computes it.

### Why is beta different on Yahoo Finance and Bloomberg?

Bloomberg's default, as a university library guide describes it, is two years of weekly returns against the S&P 500, and it shows an adjusted beta of 0.67 times the raw beta plus 0.33 next to the raw one. Yahoo labels its figure five years of monthly returns and documents nothing else. Different window, different frequency, and possibly an adjustment on one side only.

### What is adjusted beta?

A raw regression beta pulled part of the way toward 1. Marshall Blume reported in 1971 that betas measured in one period tended to be closer to 1 in the next, and adjusted betas build that tendency in. Bloomberg's weights, as a university library guide reports them, are two thirds raw and one third 1.0; Value Line says its betas are adjusted toward 1.00 without publishing the weights.

### What is the difference between levered and unlevered beta?

Levered beta is measured from the share price, so it includes the effect of the company's debt. Unlevered beta strips that out using the debt-to-equity ratio and the tax rate. Aswath Damodaran's January 2026 US sector data shows air transport at a beta of 1.19 and an unlevered beta of 0.70, at a debt-to-equity ratio of 91.17%.

### Why do Yahoo Finance and Stock Analysis show the same beta?

On 9 October 2026 both showed 1.07 for Apple, and both pages carried the same unrounded value, 1.069, which suggests a shared upstream calculation. Stock Analysis credits S&P Global Market Intelligence for its financial statistics; neither site names the source of the beta itself.

## Sources

1. [On the Assessment of Risk, The Journal of Finance 26(1), 1–10 (March 1971)](https://onlinelibrary.wiley.com/doi/10.1111/j.1540-6261.1971.tb00584.x) — Marshall E. Blume, The Journal of Finance (Wiley), read 2026-10-09
2. [Risk measurement when shares are subject to infrequent trading, Journal of Financial Economics 7(2), 197–226 (June 1979)](https://doi.org/10.1016/0304-405X(79)90013-8) — Elroy Dimson, Journal of Financial Economics (Elsevier), read 2026-10-09
3. [Apple Inc. (AAPL), statistics](https://finance.yahoo.com/quote/AAPL/key-statistics/) — Yahoo Finance, read 2026-10-09
4. [The Coca-Cola Company (KO), statistics](https://finance.yahoo.com/quote/KO/key-statistics/) — Yahoo Finance, read 2026-10-09
5. [Shell plc (SHEL.L), statistics](https://finance.yahoo.com/quote/SHEL.L/key-statistics/) — Yahoo Finance, read 2026-10-09
6. [Shell plc (SHEL), statistics](https://finance.yahoo.com/quote/SHEL/key-statistics/) — Yahoo Finance, read 2026-10-09
7. [Koyfin data dictionary](https://www.koyfin.com/help/koyfin-data-dictionary/) — Koyfin, 2026-07-01
8. [Apple (AAPL) statistics and valuation](https://stockanalysis.com/stocks/aapl/statistics/) — Stock Analysis, read 2026-10-09
9. [Help, screener](https://finviz.com/help/screener.ashx) — Finviz, read 2026-10-09
10. [Company Profile Data API](https://site.financialmodelingprep.com/developer/docs/stable/profile-symbol) — Financial Modeling Prep, read 2026-10-09
11. [Investment education glossary, Beta](https://www.valueline.com/investment-education/glossary) — Value Line, read 2026-10-09
12. [Variable definitions](https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/variable.htm) — Aswath Damodaran, NYU Stern School of Business, read 2026-10-09
13. [Betas by sector (US), data as of January 2026](https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/Betas.html) — Aswath Damodaran, NYU Stern School of Business, read 2026-10-09
14. [Trader Workstation user guide, portfolio beta weighting](https://ibkrguides.com/traderworkstation/portfolio-beta-weighting.htm) — Interactive Brokers, 2025-10-08
15. [Bloomberg research guide, Beta](https://guides.lib.byu.edu/bloomberg/Beta) — Brigham Young University Library, 2026-04-24
16. [Research glossary](https://www.fidelity.com/quick-content/etf/help/research/learn_er_glossary_1.shtml) — Fidelity Investments, read 2026-10-09

*Last updated 2026-10-09. A reference page, corrected in place — not a dated post.*