How to Calculate Portfolio Beta (With a Worked Example)

Portfolio beta is the weighted average of your individual holdings' betas: multiply each position's beta by its portfolio weight and add the results up. A portfolio beta of 1.2 means that, historically, a 1% market move came with roughly a 1.2% move in your portfolio — in both directions.

What is beta, exactly?

Beta measures an asset’s estimated return sensitivity to a chosen market benchmark. R² measures the share of return variation explained by that relationship. Technically it's the covariance of the asset's returns with the market's returns, divided by the variance of the market's returns. In plain terms:

How do I calculate my portfolio's beta?

Portfolio beta = Σ (position weight × position beta) — a weighted average.

Illustrative inputs, not current stock betas:

In a 10% market decline, this portfolio would be expected to fall roughly 12.3% — before any stock-specific news, which beta deliberately ignores.

What does a calculation with sourced betas look like?

The same four-position sample used in the interactive homepage example has a weighted beta of 1.22. These are the cached provider betas retrieved for the example on September 8, 2026, combined using our beta methodology. They are a dated input snapshot; provider windows and estimation methods can differ.

Download the supplied beta inputs and contributions (CSV). Weights are fractions of the portfolio; beta and weighted contributions are unitless. The snapshot does not retain each provider beta’s benchmark, estimation window or return frequency, so this file reproduces the weighted arithmetic only. The price-history dates stored elsewhere in the example are not evidence of the provider’s beta window.

The calculation uses all four positions. It is a weighted average of the supplied betas, not a fresh regression on the combined portfolio. VGT is 20% of the allocation but contributes 0.336 of its 1.219 beta. The calculation makes that contribution visible without suggesting a change to the allocation.

Calculate portfolio beta from your own holdings by entering tickers and share counts. The free scan shows the result for up to 20 holdings without an account. All investing involves risk, including possible loss of principal, and past performance does not guarantee future results.

Which benchmark should beta be measured against?

Beta is a two-argument function, and the second argument is usually left unstated. There is no such thing as "the beta of VXUS" — only its beta against a named benchmark, over a named window, at a named return frequency.

Three choices decide the number:

• The benchmark. Beta against a total-market index and beta against the S&P 500 are different quantities. For an international fund the choice matters enormously, because a US benchmark cannot explain the part of the fund that isn't US.
• The window. A five-year daily beta and a one-year daily beta over the same fund routinely differ by 0.1 or more, and the difference is real information about regime change, not error.
• The frequency. Daily, weekly and monthly returns give different betas for the same pair of series. Longer differencing intervals damp the effect of non-synchronous closes and raise the measured beta for assets that trade on a different clock.

PortLens starts with the beta published in each holding's provider profile: Financial Modeling Prep for US listings and Yahoo Finance for international listings. When Yahoo has no figure, PortLens computes beta from price history as covariance with the S&P 500 total-return index divided by market variance, requiring at least 30 return observations. Crypto beta is always computed from price history with weekly sampling — the full recipe is in our beta methodology.

Why do two sources report different betas for the same fund?

Because of the three choices above, mostly. It is worth seeing how large the gap can be.

The table below puts the beta published on each fund's PortLens page next to the market loading from an independent regression: the fund's daily excess returns against the Fama-French daily market factor, over the 1,206 trading days from August 2021 to May 2026.

The broad US index funds agree to within 0.04. The disagreement grows as the fund moves away from the benchmark it is being measured against — and it is largest for VXUS, at 1.07 against 0.74, where the two estimates are asking genuinely different questions of a fund that is not a US fund.

None of these numbers is wrong. They are answers to differently specified questions, which is the practical reason to care what your tool measured against rather than treating beta as a property of the asset. A beta quoted without its benchmark, window and frequency is a number you cannot reproduce.

What details make naive beta calculations wrong?

Getting a usable number means handling four edge cases most spreadsheets skip:

• The benchmark should include dividends. Beta against a price-only index (^GSPC) understates the benchmark's return by 1–2% a year. Use a total-return series (^SP500TR).
• Partially missing betas shouldn't default to 1.0. New listings and thinly traded assets often lack enough history for a meaningful beta (a common minimum is ~30 return observations). When at least one holding has a usable beta, PortLens excludes the missing holdings and renormalizes the remaining weights. If no holding has a usable beta, the portfolio calculation falls back to the market default of 1.0 rather than failing entirely.
• ETFs need a real beta, not a guess. A fund's beta reflects its underlying holdings — an S&P 500 fund sits near 1.0, a tech sector fund well above it. (What a fund actually holds is checkable on our ETF overlap pages.)
• Crypto betas need matched timestamps. Crypto trades 24/7 while equities close at 4pm ET; computing daily-return beta across mismatched closes biases it toward zero. Weekly sampling fixes the mismatch.

The renormalization rule deserves a worked case, because it is the one that silently changes an answer. For a separate hypothetical case, take weights of 50%, 30% and 20%, with betas of 1.02, 1.07 and an unavailable value. Defaulting it to 1.0 gives 0.51 + 0.321 + 0.200 = 1.03. Excluding it and renormalizing the remaining 80% gives (0.5 × 1.02 + 0.3 × 1.07) ÷ 0.8 = 1.04. Both are close here because the missing holding was equity-like. Replace it with a cash-like position that genuinely has no market sensitivity and the default answer is off by the whole difference. That is why "assume 1.0" fails in exactly the case where it matters.

The scan follows the provider precedence described above, applies the exclusion and all-missing rules to unavailable betas, and uses the total-return benchmark and weekly sampling when it computes beta from price history. The full recipe, including the benchmark and data sources, is in our beta methodology.

What is beta actually good for?

1. Sizing your downside. Beta × expected market drawdown is a first-order estimate of your systematic loss.
2. Risk-adjusting your returns. Metrics like the Treynor ratio (excess return ÷ beta) and Jensen's alpha (return above what your beta predicts) separate skill from simply holding a high-beta portfolio.
3. Spotting accidental leverage. A "balanced" portfolio with a computed beta of 1.4 is telling you something the holdings list doesn't.

For a hypothetical 20% market decline, multiplying by the sample's rounded beta of 1.22 gives −24.4%. That is the beta-only scenario calculation. It excludes company-specific events and changes in the relationship with the market, so it cannot establish how much this portfolio would actually lose.

How much of my risk does beta actually describe?

Less than the single number suggests, and the amount varies enormously by holding. R² answers this directly: it is the share of a holding's daily movement that the market explains.

For the broad index funds above, R² runs 99.0% to 99.7% — beta is very nearly the whole story. For VXUS it is 65.5%, and for a single stock it collapses: over the same window, Coca-Cola's daily returns regress on the market factor with a beta of 0.25 and an R² of 7.8%. More than nine-tenths of what that stock did had nothing to do with the market at all.

This is the structural reason beta is a portfolio measure that behaves badly on concentrated portfolios. The more of your money sits in a few names, the more of your risk lives in the part beta does not model — and a portfolio's beta will look reassuringly moderate right up until one holding moves on its own news.

What are beta's limitations?

Beta is backward-looking, assumes the relationship with the market is stable, and says nothing about company-specific risk (a beta-0.9 stock can still fall 60% on its own news). Correlations also converge in crises, so low-beta portfolios drop more in crashes than their beta implies. Use it as one lens alongside diversification measures like effective holdings.

Three more edges worth stating plainly:

• It is a measurement, not a property. Change the benchmark, window or frequency and the number changes, as the table above shows.
• It is silent about what you own. Two portfolios with identical betas can hold entirely different companies — beta cannot see the concentration that ETF overlap and look-through measures are built to find.
• It is symmetric by construction. A beta of 1.2 describes amplified moves in both directions. Nothing in the calculation distinguishes upside from downside, which is why it is a poor standalone answer to "how much could I lose".

Key takeaways

• Portfolio beta is a weighted average: Σ (weight × beta). Nothing more complicated is needed once you have honest inputs.
• The inputs are the hard part. Beta needs a stated benchmark, window and return frequency; without them it is not reproducible.
• The September 8 example combines four sourced betas into 1.22; each holding contributes its weight multiplied by its beta.
• When only some holdings lack beta, exclude them and renormalize the rest. Only an all-missing portfolio falls back to 1.0 so the calculation can complete.
• Check R² before trusting beta. At 99% it describes an index fund; at 15% it barely describes a single stock.

Revision note, September 14, 2026: added downloadable weighted-beta inputs and their provenance limits, and distinguished beta from R². The retrieved inputs remain dated September 8.

This article is for information and education only and is not investment advice. See our methodology and disclosures.