Direct answer: The evidence supports diversification as a risk-reduction tool but with important limitations: it reduces idiosyncratic (single-stock) risk effectively, but correlation between assets rises toward 1 in severe market stress, reducing the benefit of diversification precisely when it would be most valuable. Diversification reduces volatility and manages specific failure risks; it does not eliminate the risk of a broad market decline.
What the Evidence Says About Diversification
Key Takeaways
- Holding 15 to 20 randomly selected stocks captures roughly 90% of the diversification benefit available from a full market portfolio in normal conditions; moving from 1 to 10 stocks reduces standard deviation dramatically, from 10 to 20 adds less, and beyond 30 the marginal benefit is minimal.
- In the 2008 financial crisis and March 2020, correlation across equity sectors and geographies rose sharply, reducing the realized diversification benefit at exactly the moment investors needed it most.
- Diversification across uncorrelated asset classes (stocks, bonds, real assets, commodities) provides more durable protection than diversification within a single asset class.
- Geographic diversification has historically provided meaningful risk reduction over long horizons but has failed during global crises (2008, 2020) when all equity markets fell simultaneously.
- Factor diversification (adding value, small-cap, quality, and momentum tilts to a cap-weight portfolio) is supported by evidence as an enhancement to traditional diversification.
The Idiosyncratic Risk Evidence
The foundational evidence comes from Statman (1987) and Evans and Archer (1968), who showed that random stock portfolios achieve most of the variance-reducing benefit of diversification with 15 to 20 stocks. The intuition: each stock has company-specific (idiosyncratic) risk and market-wide (systematic) risk. Holding more stocks averages out the idiosyncratic component; the market-wide component cannot be diversified within equities. Statman updated the analysis in 2004 and found that 300+ stocks are needed to match the volatility of a mutual fund, because real portfolios are not randomly selected and individual stocks cluster by sector and style.
Correlation Dynamics in Stress Periods
The critical limitation is correlation instability: assets that are weakly correlated in normal periods become highly correlated in tail events. During the 2008 crisis, correlations between U.S. equities, international equities, and credit all rose above 0.9; nearly every asset class fell together except U.S. Treasury bonds. Research by Longin and Solnik (2001) found that international equity correlations rise specifically in bear markets and fall in bull markets, the opposite of what diversification-seeking investors want. This 'correlation breakdown' is the primary reason diversification protects well against idiosyncratic and sector risks but provides much weaker protection against systemic events.
International Diversification
The evidence on international diversification is mixed. Over long periods (30+ years), international diversification has reduced portfolio volatility for U.S. investors. However, periods of maximum benefit (lower correlations) coincide with periods when international markets are outperforming; in global downturns, correlations rise and international exposure amplifies rather than buffers losses. Aswath Damodaran's analysis of U.S. versus global portfolios shows that an investor who added international exposure in the 2010s underperformed a pure U.S. portfolio due to U.S. outperformance, while an investor who added it in the 2000s significantly outperformed due to U.S. underperformance. The diversification benefit is real over complete cycles but regime-dependent.
What Diversification Cannot Do
Diversification cannot eliminate systematic market risk, which is the risk that a broad market decline reduces the value of all equities simultaneously. In 2008, a diversified portfolio across 30 domestic sectors, 20 international markets, and 10 asset classes still fell 35% to 45% for equity-heavy allocations. Diversification also cannot compensate for overpaying; a portfolio of 50 overvalued equities is not protected by its breadth. The evidence supports diversification as the most reliable free lunch in investing (reducing risk without necessarily reducing expected return) within its scope, which is idiosyncratic and sector-specific risk.
Frequently Asked Questions
How many stocks do I actually need for a diversified portfolio?
For equal-weighted random selections, 15 to 20 stocks capture roughly 90% of the diversification benefit versus a full market portfolio. In practice, because investors make non-random selections (often overweighting familiar names, recent outperformers, or specific sectors), a larger number or a passive index fund is typically needed to approach random-selection diversification. A total market index fund holds thousands of stocks; a sector ETF holds tens to hundreds within one industry.
Does adding bonds actually diversify an equity portfolio?
Historically yes, with a critical caveat. Treasury bonds were negatively correlated with equities during the 2000 and 2008 crises, rising as stocks fell. In 2022, both stocks and bonds fell simultaneously (positive correlation) as the Fed raised rates; the 60/40 portfolio fell approximately 16% in a year when many investors expected bond holdings to offset equity losses. The bond-equity correlation is generally negative over long periods but can turn positive in inflationary rate-hiking environments.
Is international diversification worth the currency risk?
The evidence suggests that over full market cycles, international diversification reduces total portfolio volatility despite adding currency risk, because currency returns are relatively uncorrelated with equity returns and add a small independent risk-reduction benefit. Many international funds hedge currency exposure; the research is mixed on whether currency-hedged or unhedged international performs better, as currency provides diversification in some regimes and drag in others.