How a Single Factor Explains Most of a Stock's Risk
Investors often treat a stock's volatility as one undifferentiated number, which makes it impossible to know how much of that risk diversification can actually remove. Splitting return into a market-driven piece and a firm-specific piece answers that question directly, and the split changes everything about how a portfolio should be built.
The core idea: two sources of return
A single-factor model treats the broad market as the one dominant force behind stock price movements and writes every stock's return as a function of that one shared factor plus everything left over that is unique to the company: R_i = α_i + β_i × R_m + e_i. Here α_i is the stock's expected return unrelated to the market, β_i measures how sensitively the stock responds to market-wide moves, R_m is the market's return, and e_i is the firm-specific residual, everything from a product recall to a surprise earnings beat that has nothing to do with the broad market that day.
The critical assumption that makes this useful is that the residual terms of different stocks are uncorrelated with one another, meaning company A's unexpected news is statistically unrelated to company B's unexpected news, once each has been stripped of its shared response to the market. Under that assumption, every stock's total risk splits cleanly into two additive pieces: systematic risk, the portion driven by the market factor, and firm-specific risk, the portion unique to the company. That split is what tells you, precisely, how much of a stock's risk diversification can remove and how much it cannot.
The math: decomposing one stock's variance
Total variance under the single-factor model is Var(R_i) = β_i² × Var(R_m) + Var(e_i), the sum of systematic variance and firm-specific variance, since the two components are uncorrelated by assumption. Take a stock with a beta of 1.3, a market standard deviation of 16% (variance 0.16² = 0.0256), and a firm-specific standard deviation of 25% (variance 0.25² = 0.0625).
Systematic variance is 1.3² × 0.0256 = 1.69 × 0.0256 = 0.0433. Total variance is 0.0433 + 0.0625 = 0.1058, giving a total standard deviation of √0.1058 = 32.5%. The fraction of this stock's variance explained by the market, called R-squared, is 0.0433 / 0.1058 = 40.9%. Despite a beta well above 1.0, meaning this stock is more sensitive to the market than average, the majority of its total risk, nearly 59%, is firm-specific and has nothing to do with market direction at all.
It is worth contrasting that 32.5% figure against a second stock with a much lower beta but higher firm-specific risk, since the two components do not move together. Take a stock with a beta of just 0.6, the same 16% market standard deviation, but a larger firm-specific standard deviation of 35% (variance 0.1225), typical of a smaller, less-followed company. Systematic variance is 0.6² × 0.0256 = 0.36 × 0.0256 = 0.00922, and total variance is 0.00922 + 0.1225 = 0.1317, a total standard deviation of √0.1317 = 36.3%, actually higher than the first, high-beta stock's 32.5% despite a beta less than half as large. Its R-squared is only 0.00922 / 0.1317 = 7.0%. A low beta is often read casually as "low risk," but this comparison shows a low-beta stock can carry more total risk than a high-beta one, once firm-specific variance is properly weighed rather than ignored.
A second example: what diversification actually removes
Now build an equally weighted portfolio of 20 stocks, each with the same beta of 1.3 and the same firm-specific variance of 0.0625, and assume their firm-specific residuals are genuinely independent of one another, as the model assumes. The portfolio's beta is the weighted average of the individual betas, which, since all 20 are identical, is still 1.3. The portfolio's systematic variance is therefore unchanged: 1.3² × 0.0256 = 0.0433.
The firm-specific variance, however, shrinks dramatically, because independent residuals partly cancel when averaged: portfolio firm-specific variance is (1/N) × average firm-specific variance = 0.0625 / 20 = 0.00313. Total portfolio variance is 0.0433 + 0.00313 = 0.0464, a standard deviation of √0.0464 = 21.5%, down from 32.5% for a single stock. The portfolio's R-squared has jumped to 0.0433 / 0.0464 = 93.3%, meaning firm-specific risk, which made up 59% of a single stock's variance, has fallen to just 6.7% of the portfolio's variance. Diversification did not touch the systematic piece at all; it erased almost all of the firm-specific piece, exactly as the model predicts.
Pushing the same logic to 100 stocks instead of 20 shrinks firm-specific variance further, to 0.0625 / 100 = 0.000625, total variance to 0.0433 + 0.000625 = 0.0439, and standard deviation to 20.95%, barely above the pure systematic figure of √0.0433 = 20.8%. Beyond roughly 20 to 30 holdings, each additional stock removes progressively less firm-specific risk, since there is less of it left to remove.
What the data shows about how much beta explains
Empirical single-factor regressions on individual U.S. stocks, using a broad index as the market proxy, typically produce R-squared figures in the range of roughly 20% to 45% for individual companies over multi-year windows, meaning the toy example above, at 40.9%, sits near the higher end of what is realistic for a single stock; many smaller or more idiosyncratic companies show R-squared figures well under 30%. Diversified sector or style portfolios, by contrast, regularly show R-squared figures above 80%, and a broad diversified equity fund tracking hundreds of holdings can show an R-squared versus a total market index approaching 95% to 99%, since almost all firm-specific noise has been averaged away by that point, leaving mostly systematic exposure.
This pattern has held up consistently across market regimes, though the exact split shifts somewhat during periods of unusually high macro-driven volatility, such as a systemic financial crisis, when nearly every stock's residual becomes temporarily more correlated with the market and with each other, pushing measured R-squared figures higher across the board for a time, before reverting toward more typical levels once conditions normalize.
Sector matters here as much as company size. Utility and consumer staples stocks, whose revenues depend less on the broader business cycle, have historically shown both lower betas, often in a 0.4 to 0.7 range against a broad market index, and comparatively higher R-squared figures than their beta alone would suggest, since what market sensitivity they do have tends to be a fairly stable, dominant share of their overall variance. Smaller technology or biotechnology companies, by contrast, often show betas above 1.2 alongside R-squared figures well under 30%, because company-specific catalysts, a drug trial result, a product launch, a single large customer contract, drive a disproportionate share of their return variation regardless of what the broad market is doing on a given day. Two stocks with the same beta can therefore carry very different total risk profiles depending on which sector, and which R-squared, sits behind that shared beta number.
Using this decomposition in a real portfolio
The practical payoff of this framework is a clear answer to a question investors ask constantly without quite framing it this way: how many individual stocks do I actually need to hold before adding one more stops meaningfully reducing my risk? The math above shows the answer directly, firm-specific variance shrinks in proportion to 1/N, so the marginal risk reduction from holding stock number 25 instead of stopping at 20 is small, while the marginal risk reduction from holding stock number 5 instead of stopping at 2 is enormous. A concentrated portfolio of 5 to 10 individual names still carries substantial, uncompensated firm-specific risk that a broad fund would have already diversified away.
The decomposition also clarifies what an investor is actually being paid for. Systematic risk, the beta-driven piece, is compensated with an expected risk premium, because it cannot be diversified away and every investor holding the market bears some version of it. Firm-specific risk carries no such guaranteed compensation, since a diversified investor could have avoided it entirely at no cost; taking it on through concentrated stock picking is a bet that a specific alpha exists, not a structurally rewarded risk. This is also the conceptual bridge to how expected returns are priced across the whole market: if firm-specific risk earned no reward on average, while systematic risk did, then two stocks with identical betas should command roughly similar expected returns despite very different total volatility, since the market has no structural reason to compensate the diversifiable portion of either one's variance.
Actionable breakdown
- Separate a stock's beta from its total risk before judging it.
- Check R-squared, not just beta, to see how much is market-driven.
- A high beta does not mean most of the risk is systematic.
- Recognize where the diversification payoff runs out.
- Most firm-specific risk reduction happens by roughly 20 to 30 holdings.
- Additional holdings beyond that mainly reduce an already small residual.
- Distinguish compensated risk from uncompensated risk.
- Systematic risk carries an expected long-run reward.
- Firm-specific risk is a bet on skill, not a structurally rewarded exposure.
- Use broad funds as the default, concentrated bets as a deliberate exception.
- A diversified fund captures systematic exposure at low cost.
- Reserve concentrated positions for genuine, identifiable conviction.
Common pitfalls
The most common pitfall is treating beta as a complete description of a stock's risk, when it only describes the systematic slice; two stocks with identical betas can have very different total volatility once firm-specific risk is included, and a low-beta stock is not automatically a low-total-risk stock if its residual variance is large.
A second pitfall is assuming residuals stay uncorrelated under all conditions. The model's diversification benefit depends on that independence assumption holding, and it weakens noticeably during systemic stress, when firm-specific news across many companies starts moving together, temporarily reducing the real-world diversification benefit below what the model implies in calmer periods.
A third pitfall is over-diversifying past the point of any real benefit, spreading a portfolio across hundreds of individual names in an attempt to eliminate firm-specific risk that a much smaller, low-cost broad fund would already have removed, while incurring far more monitoring effort and transaction cost for a negligible further reduction in variance. A fourth pitfall is comparing two stocks purely by beta when deciding which one better fits a diversified portfolio, without checking whether one of them carries substantially more uncompensated firm-specific risk than the other, since that unpriced risk does not show up in the beta figure at all and can leave an investor bearing volatility that a broader fund would have absorbed and eliminated at no additional cost.
The bottom line
A stock's risk splits cleanly into a market-driven piece that diversification cannot remove and cannot avoid, and a firm-specific piece that diversification removes almost entirely by around 20 to 30 holdings.
Related reading: stock analysis fundamentals, understanding portfolio risk, diversification and portfolio risk, the single-index model in practice, the capital asset pricing model.